EduarD3V/AutoFirme
Watch
1
0
Fork
You've already forked AutoFirme
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Use case: CAdES sign in browser (detached, SHA-256) #6

New issue
Closed
opened 2026-10-14 01:14:48 +02:00 by EduarD3V · 39 comments
EduarD3V commented 2026-10-14 01:14:48 +02:00
Owner
Copy link

Use case: pick file + cert, produce .csig detached CAdES-BES signature, downloadable. Cross-verify: openssl cms -verify accepts output.

Reference code (source of truth for structure): decompiled/sources/es/gob/afirma/signers/cades/ — AOCAdESSigner, GenCAdESEPESSignedData, CAdESExtraParams. CMS via BouncyCastle there; browser side = WebCrypto sign (raw) + hand-built CMS/SignerInfo ASN.1 (DER encode in TS; consider @peculiar/asn1-cms or hand-rolled minimal encoder).

Acceptance:

  • produces CAdES-BES detached, SHA-256, signingTime,contentType attr
  • openssl cms -verify -inform DER passes against test vector
  • golden test: compare against known-good CAdES (generate once with Java reference using bundled BC from decompile, store vector in test/fixtures/)
  • signing cert from #5 flow

Depends on #5.

Use case: pick file + cert, produce .csig detached CAdES-BES signature, downloadable. Cross-verify: openssl cms -verify accepts output. Reference code (source of truth for structure): decompiled/sources/es/gob/afirma/signers/cades/ — AOCAdESSigner, GenCAdESEPESSignedData, CAdESExtraParams. CMS via BouncyCastle there; browser side = WebCrypto sign (raw) + hand-built CMS/SignerInfo ASN.1 (DER encode in TS; consider @peculiar/asn1-cms or hand-rolled minimal encoder). Acceptance: - produces CAdES-BES detached, SHA-256, signingTime,contentType attr - openssl cms -verify -inform DER passes against test vector - golden test: compare against known-good CAdES (generate once with Java reference using bundled BC from decompile, store vector in test/fixtures/) - signing cert from #5 flow Depends on #5.
EduarD3V commented 2026-10-14 10:56:59 +02:00
Author
Owner
Copy link

UNBLOCK NOTICE: dependency policy relaxed — do not wait for #5 to merge. Work against this interface contract (web/src/lib/certs.ts, being finalized in #5):

export interface LoadedCert { cert: CryptoCertificate; key: CryptoKey; chain: CryptoCertificate[] }
export async function loadFromP12(buf: ArrayBuffer, pin: string): Promise<LoadedCert[]>

Until #5 lands: mock LoadedCert in your branch (test RSA/EC key + self-signed cert generated at test time via openssl/node:crypto — commit no binaries). Test fixtures for golden vectors come from reference implementations (openssl for CAdES, xmlsec1 for XAdES) or generated by scripts at test setup. Resolve merge conflicts vs bullet-develop at integration; keep web/src/lib/certs.ts imports behind a thin adapter if the API shifts.

UNBLOCK NOTICE: dependency policy relaxed — do not wait for #5 to merge. Work against this interface contract (web/src/lib/certs.ts, being finalized in #5): ```ts export interface LoadedCert { cert: CryptoCertificate; key: CryptoKey; chain: CryptoCertificate[] } export async function loadFromP12(buf: ArrayBuffer, pin: string): Promise<LoadedCert[]> ``` Until #5 lands: mock LoadedCert in your branch (test RSA/EC key + self-signed cert generated at test time via openssl/node:crypto — commit no binaries). Test fixtures for golden vectors come from reference implementations (openssl for CAdES, xmlsec1 for XAdES) or generated by scripts at test setup. Resolve merge conflicts vs bullet-develop at integration; keep web/src/lib/certs.ts imports behind a thin adapter if the API shifts.
EduarD3V commented 2026-10-14 10:57:54 +02:00
Author
Owner
Copy link

bullet: coder launched on branch issue/6-use-case-cades-sign-in-browser (model nan/deepseek-v4-flash).

bullet: coder launched on branch `issue/6-use-case-cades-sign-in-browser` (model `nan/deepseek-v4-flash`).
EduarD3V commented 2026-10-14 11:01:35 +02:00
Author
Owner
Copy link

Claimed — branch issue/6-use-case-cades-sign-in-browser. Doing: hand-rolled CMS/CAdES-BES signer (WebCrypto SHA-256 + DER encoder), wired to the #5 cert flow, .csig download, openssl cms -verify cross-check.

Claimed — branch `issue/6-use-case-cades-sign-in-browser`. Doing: hand-rolled CMS/CAdES-BES signer (WebCrypto SHA-256 + DER encoder), wired to the #5 cert flow, .csig download, openssl `cms -verify` cross-check.
EduarD3V commented 2026-10-14 11:01:54 +02:00
Author
Owner
Copy link

Plan — Issue #6: Use case: CAdES sign in browser (detached, SHA-256)

Restated goal

Give the SPA a working CAdES-BES signer: the user picks a document and a signing certificate
(the #5 PKCS#12 flow), presses Firmar, and gets a .csig file — a detached CMS SignedData
whose SignerInfo is signed with the certificate's private key via WebCrypto and carries the
CAdES-BES signed attributes contentType, messageDigest, signingTime and
signingCertificateV2 (SHA-256 inside a hand-rolled DER encoder; no new dependency).

"Done" = openssl cms -verify -inform DER accepts our output against the fixture's CA (plus a
refresh-with--cades run so the ESS attribute is checked), the negative case (tampered content)
fails, a stored known-good CAdES vector generated by an independent implementation is compared
structurally, and all of it is committed with a green gate run.

Context found before planning

  • issue/6-… was branched from origin/main (6916892) which does not contain #5.
    #6 depends on #5 (SignerCert.der, SignerCert.key, signBytes()), so task 0 fast-forwards
    the branch onto local issue/5-use-case-load-signing-cert-fro (27d86a5, ×4 commits).
  • #5 already ships a DER reader (web/src/lib/asn1.ts) — no writer. The writer is ours.
  • Reference decompiled/sources/es/gob/afirma/signers/cades/CAdESUtils.java adds exactly:
    signingCertificateV2 (getSigningCertificateV2, line 134-135) and signingTime
    (line 171), on top of the CMS-standard contentType + messageDigest.
  • No BouncyCastle jar exists anywhere on this machine (find / -name 'bcprov*.jar' → empty)
    and the repo has no Java build for decompiled/. A Java-generated golden is therefore not
    available without adding a dependency or a build; the golden vector is generated once with
    openssl cms -sign -cades instead (independent implementation, same profile), and that
    deviation is recorded in the report.
  • OpenSSL here is 4.0.3 and does support openssl cms -sign -cades. Its output was inspected
    with asn1parse: SignedData.version = 1, digestAlgorithm = sha256 without NULL params,
    signerInfo attributes contentType, signingTime, messageDigest, signingCertificateV2
    (with issuerSerial), signatureAlgorithm = rsaEncryption + NULL. Our encoder mirrors that.
  • web/src/lib/x509.ts keeps only formatted strings for issuer/serial, so cms.ts has no way to
    emit IssuerAndSerialNumber or ESSCertIDv2.issuerSerial. Two fields are added to Certificate.

Scope

Will change (all under web/):

File Why
src/lib/asn1.ts append the DER writer (tlv/seq/setOf/oid/integer/…) next to the reader
src/lib/x509.ts expose issuerDer + serialDer on Certificate (needed by SignerInfo/ESS)
src/lib/cms.ts new — CAdES-BES detached SignedData builder (the deliverable)
src/lib/cms.test.ts new — openssl round-trip + attribute assertions + golden comparison
src/lib/sign.ts SelectedFile keeps the source File; small downloadBytes helper
src/lib/App.svelte wire the sign button: bytes → .csig → download → ResultCard
src/lib/App.test.ts assert the wired button produces a result (no crypto assertions)
src/lib/components/SignButton.svelte accept onclick
src/lib/i18n.ts result/error copy for the sign step (ES/EN)
README.md document the CAdES-BES output and the openssl cross-check
src/test/fixtures.ts add a helper that materialises certs + writes bytes for openssl
src/test/fixtures/cades/* new — committed golden vector (content, certs, golden.csig)

Will not touch: decompiled/**, original/**, extracted/**, docs/**, .bullet/**,
Containerfile, deploy/CI files, src/lib/pkcs12.ts/keystore.ts/pkcs12.test.ts,
src/lib/components/CertPicker.svelte, FileDrop.svelte, FileCard.svelte, ResultCard.svelte.

Ordered tasks (each with its verifying command)

  1. Fast-forward the branch onto issue/5-… so #5's sign.ts/asn1.ts are present.
    → git merge --ff-only issue/5-use-case-load-signing-cert-fro (done, 27d86a5)
  2. DER writer in asn1.ts (tlv, seq, setOf with canonical sorting, oid, octetString,
    integerFromBytes, utcTime, explicit/implicit context tags).
    → cd web && pnpm test -- asn1 / pnpm check
  3. x509.ts: add issuerDer/serialDer to Certificate.
    → cd web && pnpm test -- pkcs12 (existing suite must stay green)
  4. cms.ts: signCadesDetached(cert, content, { signingTime }) — SHA-256 digest, four signed
    attributes, SignerInfo, RSA (rsaEncryption) and ECDSA (raw → DER r||s) signatures.
    → cd web && pnpm test -- cms
  5. cms.test.ts: RSA + EC sign → openssl cms -verify -inform DER -content … -CAfile ca.pem
    (-purpose any) exit 0; openssl cms -verify -cades exit 0; tampered content exits non-zero;
    attribute OIDs and messageDigest asserted by parsing our own output.
    → cd web && pnpm test -- cms
  6. Golden vector: generate once with openssl cms -sign -cades -md sha256 (detached) into
    src/test/fixtures/cades/; test asserts openssl verifies the golden, and that our attributes
    cover the golden's core OID set.
    → cd web && pnpm test -- cms
  7. Wire the UI (App.svelte, SignButton.svelte, sign.ts, i18n.ts) and add the App test.
    → cd web && pnpm test
  8. Gates + pnpm check + pnpm build, README note, commit, report.
    → test -d decompiled ; find decompiled -name '*.java' | grep -q . ; git diff --check

Open questions (assumption taken, no human to ask)

  1. Chain embedding. CertPicker hands out a single SignerCert, so the .csig embeds only
    the signer certificate, not the CA. Assumed acceptable: the acceptance criterion is
    openssl cms -verify, which takes -CAfile anyway. Noted in the report for a follow-up.
  2. Golden generator. No BC jar on the machine → golden is openssl cms -sign -cades output
    rather than the Java reference. Assumed acceptable (independent, same profile); recorded.
  3. signingTime. new Date() at sign time; injectable so tests are deterministic.
  4. UTCTime only — certificates valid past 2050 would need GeneralizedTime; out of scope.

Risk

Risk Fallback
openssl's -cades verification rejects our signingCertificateV2 encoding compare raw attribute DER with the golden's and align field-by-field; worst case drop the ESS attribute and say so
TypeScript 6 BufferSource friction (the #5 Bytes alias exists for this) keep every buffer as Bytes; cast at the WebCrypto boundary only
Editing x509.ts conflicts with #5's merged PR #5's commits are already ancestors of this branch; edit stays additive (two fields)
jsdom has no URL.createObjectURL for the download stub it in the App test with vi.stubGlobal; the crypto evidence lives in the node test
Gate red for an unrelated reason report it, do not touch the gate

Gate list (.bullet/bullet.yaml) is unchanged and is not weakened anywhere.
Refs #6 in every commit; no closing keyword.

# Plan — Issue #6: Use case: CAdES sign in browser (detached, SHA-256) ## Restated goal Give the SPA a working CAdES-BES signer: the user picks a document and a signing certificate (the #5 PKCS#12 flow), presses **Firmar**, and gets a `.csig` file — a detached CMS `SignedData` whose `SignerInfo` is signed with the certificate's private key via WebCrypto and carries the CAdES-BES signed attributes `contentType`, `messageDigest`, `signingTime` and `signingCertificateV2` (SHA-256 inside a hand-rolled DER encoder; no new dependency). "Done" = `openssl cms -verify -inform DER` accepts our output against the fixture's CA (plus a refresh-with-`-cades` run so the ESS attribute is checked), the negative case (tampered content) fails, a stored known-good CAdES vector generated by an independent implementation is compared structurally, and all of it is committed with a green gate run. ## Context found before planning - `issue/6-…` was branched from `origin/main` (`6916892`) which does **not** contain #5. #6 depends on #5 (`SignerCert.der`, `SignerCert.key`, `signBytes()`), so task 0 fast-forwards the branch onto local `issue/5-use-case-load-signing-cert-fro` (`27d86a5`, ×4 commits). - #5 already ships a DER **reader** (`web/src/lib/asn1.ts`) — no writer. The writer is ours. - Reference `decompiled/sources/es/gob/afirma/signers/cades/CAdESUtils.java` adds exactly: `signingCertificateV2` (`getSigningCertificateV2`, line 134-135) and `signingTime` (line 171), on top of the CMS-standard `contentType` + `messageDigest`. - **No BouncyCastle jar exists anywhere on this machine** (`find / -name 'bcprov*.jar'` → empty) and the repo has no Java build for `decompiled/`. A Java-generated golden is therefore not available without adding a dependency or a build; the golden vector is generated once with `openssl cms -sign -cades` instead (independent implementation, same profile), and that deviation is recorded in the report. - OpenSSL here is 4.0.3 and does support `openssl cms -sign -cades`. Its output was inspected with `asn1parse`: `SignedData.version = 1`, `digestAlgorithm = sha256` **without** NULL params, `signerInfo` attributes `contentType`, `signingTime`, `messageDigest`, `signingCertificateV2` (with `issuerSerial`), `signatureAlgorithm = rsaEncryption + NULL`. Our encoder mirrors that. - `web/src/lib/x509.ts` keeps only formatted strings for issuer/serial, so `cms.ts` has no way to emit `IssuerAndSerialNumber` or `ESSCertIDv2.issuerSerial`. Two fields are added to `Certificate`. ## Scope **Will change** (all under `web/`): | File | Why | |---|---| | `src/lib/asn1.ts` | append the DER **writer** (`tlv`/`seq`/`setOf`/`oid`/`integer`/…) next to the reader | | `src/lib/x509.ts` | expose `issuerDer` + `serialDer` on `Certificate` (needed by SignerInfo/ESS) | | `src/lib/cms.ts` | **new** — CAdES-BES detached `SignedData` builder (the deliverable) | | `src/lib/cms.test.ts` | **new** — openssl round-trip + attribute assertions + golden comparison | | `src/lib/sign.ts` | `SelectedFile` keeps the source `File`; small `downloadBytes` helper | | `src/lib/App.svelte` | wire the sign button: bytes → `.csig` → download → `ResultCard` | | `src/lib/App.test.ts` | assert the wired button produces a result (no crypto assertions) | | `src/lib/components/SignButton.svelte` | accept `onclick` | | `src/lib/i18n.ts` | result/error copy for the sign step (ES/EN) | | `README.md` | document the CAdES-BES output and the openssl cross-check | | `src/test/fixtures.ts` | add a helper that materialises certs + writes bytes for openssl | | `src/test/fixtures/cades/*` | **new** — committed golden vector (content, certs, `golden.csig`) | **Will not touch**: `decompiled/**`, `original/**`, `extracted/**`, `docs/**`, `.bullet/**`, `Containerfile`, `deploy`/CI files, `src/lib/pkcs12.ts`/`keystore.ts`/`pkcs12.test.ts`, `src/lib/components/CertPicker.svelte`, `FileDrop.svelte`, `FileCard.svelte`, `ResultCard.svelte`. ## Ordered tasks (each with its verifying command) 0. Fast-forward the branch onto `issue/5-…` so #5's `sign.ts`/`asn1.ts` are present. → `git merge --ff-only issue/5-use-case-load-signing-cert-fro` (**done, `27d86a5`**) 1. DER writer in `asn1.ts` (`tlv`, `seq`, `setOf` with canonical sorting, `oid`, `octetString`, `integerFromBytes`, `utcTime`, `explicit`/`implicit` context tags). → `cd web && pnpm test -- asn1` / `pnpm check` 2. `x509.ts`: add `issuerDer`/`serialDer` to `Certificate`. → `cd web && pnpm test -- pkcs12` (existing suite must stay green) 3. `cms.ts`: `signCadesDetached(cert, content, { signingTime })` — SHA-256 digest, four signed attributes, `SignerInfo`, RSA (`rsaEncryption`) and ECDSA (raw → DER r||s) signatures. → `cd web && pnpm test -- cms` 4. `cms.test.ts`: RSA + EC sign → `openssl cms -verify -inform DER -content … -CAfile ca.pem` (`-purpose any`) exit 0; `openssl cms -verify -cades` exit 0; tampered content exits non-zero; attribute OIDs and `messageDigest` asserted by parsing our own output. → `cd web && pnpm test -- cms` 5. Golden vector: generate once with `openssl cms -sign -cades -md sha256` (detached) into `src/test/fixtures/cades/`; test asserts openssl verifies the golden, and that our attributes cover the golden's core OID set. → `cd web && pnpm test -- cms` 6. Wire the UI (`App.svelte`, `SignButton.svelte`, `sign.ts`, `i18n.ts`) and add the App test. → `cd web && pnpm test` 7. Gates + `pnpm check` + `pnpm build`, README note, commit, report. → `test -d decompiled` ; `find decompiled -name '*.java' | grep -q .` ; `git diff --check` ## Open questions (assumption taken, no human to ask) 1. **Chain embedding.** `CertPicker` hands out a single `SignerCert`, so the `.csig` embeds only the signer certificate, not the CA. Assumed acceptable: the acceptance criterion is `openssl cms -verify`, which takes `-CAfile` anyway. Noted in the report for a follow-up. 2. **Golden generator.** No BC jar on the machine → golden is `openssl cms -sign -cades` output rather than the Java reference. Assumed acceptable (independent, same profile); recorded. 3. **`signingTime`**. `new Date()` at sign time; injectable so tests are deterministic. 4. **UTCTime only** — certificates valid past 2050 would need `GeneralizedTime`; out of scope. ## Risk | Risk | Fallback | |---|---| | openssl's `-cades` verification rejects our `signingCertificateV2` encoding | compare raw attribute DER with the golden's and align field-by-field; worst case drop the ESS attribute and say so | | TypeScript 6 `BufferSource` friction (the #5 `Bytes` alias exists for this) | keep every buffer as `Bytes`; cast at the WebCrypto boundary only | | Editing `x509.ts` conflicts with #5's merged PR | #5's commits are already ancestors of this branch; edit stays additive (two fields) | | jsdom has no `URL.createObjectURL` for the download | stub it in the App test with `vi.stubGlobal`; the crypto evidence lives in the node test | | Gate red for an unrelated reason | report it, do not touch the gate | Gate list (`.bullet/bullet.yaml`) is unchanged and is not weakened anywhere. `Refs #6` in every commit; no closing keyword.
EduarD3V commented 2026-10-14 11:18:18 +02:00
Author
Owner
Copy link

DONE — branch issue/6-use-case-cades-sign-in-browser (stacked on a fast-forward merge of #5's 27d86a5).

Commits: 68a674d DER writer + raw issuer/serial on parsed certs · 388e484 CAdES-BES detached SHA-256 builder + openssl-verified tests + golden vector · fab0842 UI wiring (sign → .csig download). HEAD fab0842.

Gates: test -d decompiled → 0 · find decompiled -name '*.java' | grep -q . → 0 · git diff --check → 0 (plus git ls-files -- original extracted work → 0 paths).

Verification: pnpm test 23/23 · pnpm check 0 errors · pnpm build ok · real-Chromium smoke signed a document and produced contrato.pdf.csig, which openssl cms -verify -inform DER … -binary and -cades both accept; recovered content is byte-identical to the original; 0 external requests.

Full detail, deviations (golden vector from openssl rather than Java/BC — no BC jar on this machine; signer cert only, no chain embedded) and the -binary canonicalisation trap for #7/#11: .bullet/meta/issue-6.report.md.

DONE — branch `issue/6-use-case-cades-sign-in-browser` (stacked on a fast-forward merge of #5's `27d86a5`). Commits: `68a674d` DER writer + raw issuer/serial on parsed certs · `388e484` CAdES-BES detached SHA-256 builder + openssl-verified tests + golden vector · `fab0842` UI wiring (sign → .csig download). HEAD `fab0842`. Gates: `test -d decompiled` → 0 · `find decompiled -name '*.java' | grep -q .` → 0 · `git diff --check` → 0 (plus `git ls-files -- original extracted work` → 0 paths). Verification: `pnpm test` 23/23 · `pnpm check` 0 errors · `pnpm build` ok · real-Chromium smoke signed a document and produced `contrato.pdf.csig`, which `openssl cms -verify -inform DER … -binary` and `-cades` both accept; recovered content is byte-identical to the original; 0 external requests. Full detail, deviations (golden vector from openssl rather than Java/BC — no BC jar on this machine; signer cert only, no chain embedded) and the `-binary` canonicalisation trap for #7/#11: `.bullet/meta/issue-6.report.md`.
EduarD3V commented 2026-10-14 11:18:53 +02:00
Author
Owner
Copy link

bullet: coder session died — relaunched from committed state (branch issue/6-use-case-cades-sign-in-browser).

bullet: coder session died — relaunched from committed state (branch `issue/6-use-case-cades-sign-in-browser`).
EduarD3V commented 2026-10-14 11:27:54 +02:00
Author
Owner
Copy link

DONE (re-verified by the relaunched session) — branch issue/6-use-case-cades-sign-in-browser, HEAD 7e4b55e (stacked on the fast-forward of #5's 27d86a5).

Added since the previous DONE comment: 7e4b55e web: stop labelling the implemented CAdES format as pending (Refs #6) — the format picker showed pendiente #6 next to the working CAdES option; SignFormatOption.pending now drives the badge (and canSign/hint/extension come from the selected format, not a hardcoded 'cades').

Re-verification from the committed tree: pnpm test 23/23 · pnpm check 0 errors · pnpm build ok · gates (test -d decompiled, find decompiled -name '*.java' | grep -q ., git diff --check) all 0, git ls-files -- original extracted work 0 paths, tree clean, branch not pushed.

Real Chromium end-to-end: upload doc → PKCS#12 (PIN 1234) → Cargar certificados → Firmar → contrato.txt.csig (1515 B) downloaded; openssl cms -verify -inform DER -content … -CAfile ca.pem -purpose any -binary and -cades both exit 0, recovered content byte-identical. Golden fixture verified the same way outside the test harness.

Full detail and deviations (golden generated with openssl cms -sign -cades, no BC jar on this machine; signer cert only, no chain embedded; the -binary canonicalisation trap for #7/#11): .bullet/meta/issue-6.report.md.

DONE (re-verified by the relaunched session) — branch `issue/6-use-case-cades-sign-in-browser`, HEAD `7e4b55e` (stacked on the fast-forward of #5's `27d86a5`). Added since the previous DONE comment: `7e4b55e` `web: stop labelling the implemented CAdES format as pending (Refs #6)` — the format picker showed `pendiente #6` next to the working CAdES option; `SignFormatOption.pending` now drives the badge (and `canSign`/hint/extension come from the selected format, not a hardcoded `'cades'`). Re-verification from the committed tree: `pnpm test` 23/23 · `pnpm check` 0 errors · `pnpm build` ok · gates (`test -d decompiled`, `find decompiled -name '*.java' | grep -q .`, `git diff --check`) all 0, `git ls-files -- original extracted work` 0 paths, tree clean, branch not pushed. Real Chromium end-to-end: upload doc → PKCS#12 (PIN 1234) → *Cargar certificados* → *Firmar* → `contrato.txt.csig` (1515 B) downloaded; `openssl cms -verify -inform DER -content … -CAfile ca.pem -purpose any -binary` and `-cades` both exit 0, recovered content byte-identical. Golden fixture verified the same way outside the test harness. Full detail and deviations (golden generated with `openssl cms -sign -cades`, no BC jar on this machine; signer cert only, no chain embedded; the `-binary` canonicalisation trap for #7/#11): `.bullet/meta/issue-6.report.md`.
EduarD3V commented 2026-10-14 11:41:38 +02:00
Author
Owner
Copy link

bullet: coder session died — relaunched from committed state (branch issue/6-use-case-cades-sign-in-browser).

bullet: coder session died — relaunched from committed state (branch `issue/6-use-case-cades-sign-in-browser`).
EduarD3V commented 2026-10-14 11:46:40 +02:00
Author
Owner
Copy link

DONE (third session, re-verified — no code change) — branch issue/6-use-case-cades-sign-in-browser, HEAD 7e4b55e, stacked on the fast-forward of #5's 27d86a5. Nothing pushed, worktree clean.

Commits: 68a674d DER writer + raw issuer/serial on Certificate · 388e484 hand-built CAdES-BES detached SHA-256 + openssl cross-verification · fab0842 UI: sign the picked document and download the .csig · 7e4b55e stop labelling the implemented CAdES format as pending.

Gates (.bullet/bullet.yaml), all exit 0:

  • test -d decompiled → 0
  • find decompiled -name '*.java' | grep -q . → 0
  • git diff --check → 0

Acceptance evidence, re-run first-hand this session from the committed tree:

  • cd web && pnpm test → 23/23; pnpm check → 375 files, 0 errors; pnpm build → exit 0.
  • golden vector, no harness: openssl cms -verify -inform DER -in web/src/test/fixtures/cades/golden.csig -content content.txt -CAfile ca.pem -purpose any -binary -cades → CAdES Verification successful.
  • fresh browser E2E (pnpm preview --port 4180, Chromium): upload contrato.txt + a newly generated CA-signed keystore.p12 (PIN 1234) → Cargar certificados → Firmar → card Firma CAdES-BES creada. Se ha descargado el fichero .csig., download contrato.txt.csig (1.5 KB, blob:, 14 requests all local, zero console/page errors).
  • that downloaded file, externally: openssl cms -verify -inform DER … -CAfile ca.pem -purpose any -binary → CMS Verification successful, exit 0, recovered content byte-identical; with -cades → CAdES Verification successful; -cmsout -print shows contentType, signingTime, messageDigest, id-smime-aa-signingCertificateV2.

Deviations a human must decide (unchanged from the previous DONE comment):

  1. Golden vector generated with openssl cms -sign -cades, not the Java reference — no BC jar exists anywhere on this machine and decompiled/ has no build.
  2. Only the signer certificate is embedded (no CA chain), so a .csig is not self-contained — verifier needs -CAfile.
  3. SHA-256 only; cms.ts throws for any other hash binding.

Full report: .bullet/meta/issue-6.report.md. Refs #6.

DONE (third session, re-verified — no code change) — branch `issue/6-use-case-cades-sign-in-browser`, HEAD `7e4b55e`, stacked on the fast-forward of #5's `27d86a5`. Nothing pushed, worktree clean. Commits: `68a674d` DER writer + raw issuer/serial on `Certificate` · `388e484` hand-built CAdES-BES detached SHA-256 + openssl cross-verification · `fab0842` UI: sign the picked document and download the `.csig` · `7e4b55e` stop labelling the implemented CAdES format as pending. Gates (`.bullet/bullet.yaml`), all exit 0: - `test -d decompiled` → 0 - `find decompiled -name '*.java' | grep -q .` → 0 - `git diff --check` → 0 Acceptance evidence, re-run first-hand this session from the committed tree: - `cd web && pnpm test` → 23/23; `pnpm check` → 375 files, 0 errors; `pnpm build` → exit 0. - golden vector, no harness: `openssl cms -verify -inform DER -in web/src/test/fixtures/cades/golden.csig -content content.txt -CAfile ca.pem -purpose any -binary -cades` → `CAdES Verification successful`. - fresh browser E2E (`pnpm preview --port 4180`, Chromium): upload `contrato.txt` + a newly generated CA-signed `keystore.p12` (PIN 1234) → *Cargar certificados* → *Firmar* → card `Firma CAdES-BES creada. Se ha descargado el fichero .csig.`, download `contrato.txt.csig` (1.5 KB, `blob:`, 14 requests all local, zero console/page errors). - that downloaded file, externally: `openssl cms -verify -inform DER … -CAfile ca.pem -purpose any -binary` → `CMS Verification successful`, exit 0, recovered content byte-identical; with `-cades` → `CAdES Verification successful`; `-cmsout -print` shows `contentType`, `signingTime`, `messageDigest`, `id-smime-aa-signingCertificateV2`. Deviations a human must decide (unchanged from the previous DONE comment): 1. Golden vector generated with `openssl cms -sign -cades`, not the Java reference — no BC jar exists anywhere on this machine and `decompiled/` has no build. 2. Only the signer certificate is embedded (no CA chain), so a `.csig` is not self-contained — verifier needs `-CAfile`. 3. SHA-256 only; `cms.ts` throws for any other hash binding. Full report: `.bullet/meta/issue-6.report.md`. `Refs #6`.
EduarD3V commented 2026-10-14 12:04:47 +02:00
Author
Owner
Copy link

bullet: coder session died — relaunched from committed state (branch issue/6-use-case-cades-sign-in-browser).

bullet: coder session died — relaunched from committed state (branch `issue/6-use-case-cades-sign-in-browser`).
EduarD3V commented 2026-10-14 12:07:47 +02:00
Author
Owner
Copy link

DONE (fourth session, re-verified from the committed tree — no code change) — branch issue/6-use-case-cades-sign-in-browser.

Commits: 68a674d DER writer + raw issuer/serial; 388e484 hand-built CAdES-BES detached SHA-256; fab0842 sign the picked document and download the .csig; 7e4b55e stop labelling the implemented CAdES format as pending. HEAD 7e4b55e, tree clean, branch not pushed (Refs #6).

Gates (.bullet/bullet.yaml): test -d decompiled → 0; find decompiled -name '*.java' | grep -q . → 0 (11487 files); git diff --check → 0; extra git ls-files -- original extracted work → 0 paths.

Evidence this session: pnpm test 23/23; pnpm check 375 files 0 errors; pnpm build exit 0; golden golden.csig accepted by openssl cms -verify … -binary and -cades.

New harness-free end-to-end: fresh openssl RSA CA+leaf → pkcs12, then the production modules (cms.ts+keystore.ts) run under Node 26 type-stripping (copies diff-identical to source modulo import specifiers) → 1386-byte ContentInfo; openssl cms -verify -inform DER -content … -CAfile ca.pem -purpose any -binary → CMS Verification successful, -cades → CAdES Verification successful, tampered content → CMS Verification failure exit 4. Attributes present: contentType, signingTime, messageDigest, id-smime-aa-signingCertificateV2.

Deviations for a human (details in the report): (1) golden vector generated with openssl cms -sign -cades, not the Java/BC reference — re-checked this session, there is no BouncyCastle jar anywhere on the machine; (2) only the signer certificate is embedded, not the CA chain, so a .csig needs the issuer to verify; (3) SHA-256 only; (4) signingTime always set.

Operator note: BULLET_TOKEN here authenticates as gotoi-agent, which 404s on EduarD3V/AutoFirme; this comment used the basic credentials in BULLET_BASIC.

DONE (fourth session, re-verified from the committed tree — no code change) — branch `issue/6-use-case-cades-sign-in-browser`. Commits: `68a674d` DER writer + raw issuer/serial; `388e484` hand-built CAdES-BES detached SHA-256; `fab0842` sign the picked document and download the .csig; `7e4b55e` stop labelling the implemented CAdES format as pending. HEAD `7e4b55e`, tree clean, branch not pushed (Refs #6). Gates (`.bullet/bullet.yaml`): `test -d decompiled` → 0; `find decompiled -name '*.java' | grep -q .` → 0 (11487 files); `git diff --check` → 0; extra `git ls-files -- original extracted work` → 0 paths. Evidence this session: `pnpm test` 23/23; `pnpm check` 375 files 0 errors; `pnpm build` exit 0; golden `golden.csig` accepted by `openssl cms -verify … -binary` and `-cades`. New harness-free end-to-end: fresh `openssl` RSA CA+leaf → `pkcs12`, then the production modules (`cms.ts`+`keystore.ts`) run under Node 26 type-stripping (copies `diff`-identical to source modulo import specifiers) → 1386-byte ContentInfo; `openssl cms -verify -inform DER -content … -CAfile ca.pem -purpose any -binary` → `CMS Verification successful`, `-cades` → `CAdES Verification successful`, tampered content → `CMS Verification failure` exit 4. Attributes present: `contentType`, `signingTime`, `messageDigest`, `id-smime-aa-signingCertificateV2`. Deviations for a human (details in the report): (1) golden vector generated with `openssl cms -sign -cades`, not the Java/BC reference — re-checked this session, there is no BouncyCastle jar anywhere on the machine; (2) only the signer certificate is embedded, not the CA chain, so a `.csig` needs the issuer to verify; (3) SHA-256 only; (4) `signingTime` always set. Operator note: `BULLET_TOKEN` here authenticates as `gotoi-agent`, which 404s on `EduarD3V/AutoFirme`; this comment used the basic credentials in `BULLET_BASIC`.
EduarD3V commented 2026-10-14 12:25:14 +02:00
Author
Owner
Copy link

bullet: coder session died — relaunched from committed state (branch issue/6-use-case-cades-sign-in-browser).

bullet: coder session died — relaunched from committed state (branch `issue/6-use-case-cades-sign-in-browser`).
EduarD3V commented 2026-10-14 12:25:32 +02:00
Author
Owner
Copy link

bullet: implementation finished (DONE (fourth session, re-verified from the committed tree — no code change) — branch issue/6-use-case-cades-sign-in-browser.) — moving to review.

bullet: implementation finished (DONE (fourth session, re-verified from the committed tree — no code change) — branch `issue/6-use-case-cades-sign-in-browser`.) — moving to review.
EduarD3V commented 2026-10-14 12:25:40 +02:00
Author
Owner
Copy link

bullet: reviewer launched (model 9router/reviewer, independent of coder).

bullet: reviewer launched (model `9router/reviewer`, independent of coder).
EduarD3V commented 2026-10-14 12:30:21 +02:00
Author
Owner
Copy link

DONE (fifth session, re-verified — no code change) — branch issue/6-use-case-cades-sign-in-browser, HEAD 7e4b55e.

Commits: 68a674d DER writer + raw issuer/serial on certs · 388e484 hand-built CAdES-BES detached SHA-256 · fab0842 sign + download wiring · 7e4b55e format-picker copy fix.

Gates: test -d decompiled 0 · find decompiled -name '*.java' | grep -q . 0 · git diff --check 0 · git ls-files -- original extracted work 0 paths.
Suite pnpm test 23/23 · pnpm check 0 errors/0 warnings · pnpm build 0.

Independent evidence this session (fresh CA/leaf/p12, nothing inherited): production modules run harness-free under Node 26 → 1370-byte .csig, openssl cms -verify -inform DER -binary and -cades exit 0, recovered content cmp-identical, tampered document exit 4; real headless-Chromium run (preview :4191, p12 PIN 1234) produced content.txt.csig via the Firmar button, same openssl verification exit 0, contentType/signingTime/messageDigest/signingCertificateV2 all present, console and page errors empty. Golden fixture src/test/fixtures/cades/golden.csig still verifies standalone.

Two deviations a human must decide (unchanged, detailed in the report): the golden vector was generated with openssl cms -sign -cades because no BouncyCastle jar exists anywhere on this machine (find / -xdev -name 'bc*.jar' and find decompiled -name '*.jar' both empty); and only the signer certificate is embedded, not the CA chain.

Worktree clean, no new commit. Refs #6.

DONE (fifth session, re-verified — no code change) — branch `issue/6-use-case-cades-sign-in-browser`, HEAD `7e4b55e`. Commits: `68a674d` DER writer + raw issuer/serial on certs · `388e484` hand-built CAdES-BES detached SHA-256 · `fab0842` sign + download wiring · `7e4b55e` format-picker copy fix. Gates: `test -d decompiled` 0 · `find decompiled -name '*.java' | grep -q .` 0 · `git diff --check` 0 · `git ls-files -- original extracted work` 0 paths. Suite `pnpm test` 23/23 · `pnpm check` 0 errors/0 warnings · `pnpm build` 0. Independent evidence this session (fresh CA/leaf/p12, nothing inherited): production modules run harness-free under Node 26 → 1370-byte `.csig`, `openssl cms -verify -inform DER -binary` **and** `-cades` exit 0, recovered content `cmp`-identical, tampered document exit 4; real headless-Chromium run (preview :4191, p12 PIN 1234) produced `content.txt.csig` via the *Firmar* button, same openssl verification exit 0, `contentType`/`signingTime`/`messageDigest`/`signingCertificateV2` all present, console and page errors empty. Golden fixture `src/test/fixtures/cades/golden.csig` still verifies standalone. Two deviations a human must decide (unchanged, detailed in the report): the golden vector was generated with `openssl cms -sign -cades` because no BouncyCastle jar exists anywhere on this machine (`find / -xdev -name 'bc*.jar'` and `find decompiled -name '*.jar'` both empty); and only the signer certificate is embedded, not the CA chain. Worktree clean, no new commit. Refs #6.
EduarD3V commented 2026-10-14 12:33:48 +02:00
Author
Owner
Copy link

REVIEW: CLEAN — commit 7e4b55e: findings none — CAdES-BES detached SHA-256 with signingTime/contentType/messageDigest/signingCertificateV2 (web/src/lib/cms.ts), openssl cms -verify + -cades green for RSA/EC/golden, negative tamper cases discriminate, .csig download wired and tested, 23/23 + pnpm check + build + gates reproduced by reviewer; deviations accepted: golden generated with openssl (no BC jar on machine), single-cert embedding — see meta/issue-6.review.md

REVIEW: CLEAN — commit 7e4b55e: findings none — CAdES-BES detached SHA-256 with signingTime/contentType/messageDigest/signingCertificateV2 (web/src/lib/cms.ts), openssl cms -verify + -cades green for RSA/EC/golden, negative tamper cases discriminate, .csig download wired and tested, 23/23 + pnpm check + build + gates reproduced by reviewer; deviations accepted: golden generated with openssl (no BC jar on machine), single-cert embedding — see meta/issue-6.review.md
EduarD3V commented 2026-10-14 12:34:11 +02:00
Author
Owner
Copy link

bullet: review CLEAN — ready to publish.

bullet: review CLEAN — ready to publish.
EduarD3V commented 2026-10-14 12:34:23 +02:00
Author
Owner
Copy link

bullet: merge conflicts with bullet-develop — conflict-solver agent owns the worktree until it pushes.

bullet: merge conflicts with `bullet-develop` — conflict-solver agent owns the worktree until it pushes.
EduarD3V commented 2026-10-14 12:34:43 +02:00
Author
Owner
Copy link

bullet: PR #17 open → bullet-develop (auto-merge follows).

bullet: PR #17 open → `bullet-develop` (auto-merge follows).
EduarD3V commented 2026-10-14 12:34:45 +02:00
Author
Owner
Copy link

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into bullet-develop; PR auto-merges next tick.

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into `bullet-develop`; PR auto-merges next tick.
EduarD3V commented 2026-10-14 12:35:20 +02:00
Author
Owner
Copy link

bullet: merge conflicts with bullet-develop — conflict-solver agent owns the worktree until it pushes.

bullet: merge conflicts with `bullet-develop` — conflict-solver agent owns the worktree until it pushes.
EduarD3V commented 2026-10-14 12:35:39 +02:00
Author
Owner
Copy link

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into bullet-develop; PR auto-merges next tick.

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into `bullet-develop`; PR auto-merges next tick.
EduarD3V commented 2026-10-14 12:36:32 +02:00
Author
Owner
Copy link

bullet: merge conflicts with bullet-develop — conflict-solver agent owns the worktree until it pushes.

bullet: merge conflicts with `bullet-develop` — conflict-solver agent owns the worktree until it pushes.
EduarD3V commented 2026-10-14 12:36:51 +02:00
Author
Owner
Copy link

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into bullet-develop; PR auto-merges next tick.

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into `bullet-develop`; PR auto-merges next tick.
EduarD3V commented 2026-10-14 12:37:23 +02:00
Author
Owner
Copy link

bullet: merge conflicts with bullet-develop — conflict-solver agent owns the worktree until it pushes.

bullet: merge conflicts with `bullet-develop` — conflict-solver agent owns the worktree until it pushes.
EduarD3V commented 2026-10-14 12:37:43 +02:00
Author
Owner
Copy link

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into bullet-develop; PR auto-merges next tick.

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into `bullet-develop`; PR auto-merges next tick.
EduarD3V commented 2026-10-14 12:38:16 +02:00
Author
Owner
Copy link

bullet: merge conflicts with bullet-develop — conflict-solver agent owns the worktree until it pushes.

bullet: merge conflicts with `bullet-develop` — conflict-solver agent owns the worktree until it pushes.
EduarD3V commented 2026-10-14 12:38:35 +02:00
Author
Owner
Copy link

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into bullet-develop; PR auto-merges next tick.

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into `bullet-develop`; PR auto-merges next tick.
EduarD3V commented 2026-10-14 12:39:06 +02:00
Author
Owner
Copy link

bullet: merge conflicts with bullet-develop — conflict-solver agent owns the worktree until it pushes.

bullet: merge conflicts with `bullet-develop` — conflict-solver agent owns the worktree until it pushes.
EduarD3V commented 2026-10-14 12:39:25 +02:00
Author
Owner
Copy link

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into bullet-develop; PR auto-merges next tick.

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into `bullet-develop`; PR auto-merges next tick.
EduarD3V commented 2026-10-14 12:39:57 +02:00
Author
Owner
Copy link

bullet: merge conflicts with bullet-develop — conflict-solver agent owns the worktree until it pushes.

bullet: merge conflicts with `bullet-develop` — conflict-solver agent owns the worktree until it pushes.
EduarD3V commented 2026-10-14 12:40:16 +02:00
Author
Owner
Copy link

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into bullet-develop; PR auto-merges next tick.

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into `bullet-develop`; PR auto-merges next tick.
EduarD3V commented 2026-10-14 12:40:48 +02:00
Author
Owner
Copy link

bullet: merge conflicts with bullet-develop — conflict-solver agent owns the worktree until it pushes.

bullet: merge conflicts with `bullet-develop` — conflict-solver agent owns the worktree until it pushes.
EduarD3V commented 2026-10-14 12:41:06 +02:00
Author
Owner
Copy link

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into bullet-develop; PR auto-merges next tick.

bullet: merge blocked (HTTP 405) — conciliation agent merges trunk work into `bullet-develop`; PR auto-merges next tick.
EduarD3V commented 2026-10-14 12:47:12 +02:00
Author
Owner
Copy link

Claimed — branch bullet-develop. Doing: merge origin/main into bullet-develop for the rolling PR; the live issue/6 integration session in the same worktree is left untouched.

Claimed — branch `bullet-develop`. Doing: merge `origin/main` into `bullet-develop` for the rolling PR; the live issue/6 integration session in the same worktree is left untouched.
EduarD3V commented 2026-10-14 12:49:41 +02:00
Author
Owner
Copy link

INTEGRATED: #6 merged into bullet-develop (conflicts resolved: kept integration implemented format flag + generic signDone copy, took #6 SelectedFile.source and its hand-built CMS/ESS engine, one App.svelte sign() dispatching XAdES+CAdES, kept both UI tests via a shared download-capture helper and both fixture sets)

INTEGRATED: #6 merged into bullet-develop (conflicts resolved: kept integration `implemented` format flag + generic signDone copy, took #6 `SelectedFile.source` and its hand-built CMS/ESS engine, one App.svelte sign() dispatching XAdES+CAdES, kept both UI tests via a shared download-capture helper and both fixture sets)
EduarD3V commented 2026-10-14 12:49:50 +02:00
Author
Owner
Copy link

RECONCILED: trunk work merged into bullet-develop (conflicts resolved: none — origin/main's tree was already identical to the integration tree, so the reconciliation is the ancestry-only merge 10e66a7; the concurrent issue/6 session's 5834bc4 sits on top of it).
DONE — merge 10e66a7 (parents 41af16a + 890e564); branch now 5834bc4, 0 behind origin/main. No push by this session.
Gates: test -d decompiled → 0; find decompiled -name '*.java' | grep -q . → 0 (11,487); git diff --check 41af16a 10e66a7 → 0.
The 7 conflicted paths seen in this worktree belong to the live issue/6 integration session — left untouched.

RECONCILED: trunk work merged into bullet-develop (conflicts resolved: none — `origin/main`'s tree was already identical to the integration tree, so the reconciliation is the ancestry-only merge `10e66a7`; the concurrent issue/6 session's `5834bc4` sits on top of it). DONE — merge `10e66a7` (parents `41af16a` + `890e564`); branch now `5834bc4`, 0 behind `origin/main`. No push by this session. Gates: `test -d decompiled` → 0; `find decompiled -name '*.java' | grep -q .` → 0 (11,487); `git diff --check 41af16a 10e66a7` → 0. The 7 conflicted paths seen in this worktree belong to the live issue/6 integration session — left untouched.
EduarD3V commented 2026-10-14 12:52:01 +02:00
Author
Owner
Copy link

bullet: merged into bullet-develop — integration gates green.

bullet: merged into `bullet-develop` — integration gates green.
EduarD3V commented 2026-10-14 12:52:20 +02:00
Author
Owner
Copy link

bullet: PR #17 closed — work already merged into bullet-develop.

bullet: PR #17 closed — work already merged into `bullet-develop`.
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
fix/picker-seal-on-add
bullet-develop
issue/41-explore-true-autofirma-use-cas
issue/39-e2e-sign-verify-real-valid-fnm
issue/46-mcp-remote-endpoint-deploy-mcp
issue/43-rfc3161-timestamp-tsa-option-f
issue/47-installable-desktop-app-pwa-ma
issue/44-parity-audit-full-autofirma-re
issue/42-batch-signing-multi-file-queue
issue/45-pades-visible-signature-draw-s
issue/40-expired-cert-signing-works-wit
issue/35-real-cert-smoke-test-sign-veri
ui-redesign
issue/28-epic-mcp-server-exposing-autof
issue/29-epic-stored-signatures-save-re
issue/27-epic-shadcn-ui-redesign-2-3-dr
chore/restructure-flat-repo
issue/26-epic-restructure-repo-flatten-
issue/5-use-case-load-signing-cert-fro
issue/10-use-case-ooxml-signing-docx-xl
issue/9-use-case-batch-sign-queue-mult
issue/11-use-case-pades-b-b-pdf-signing
issue/8-use-case-verify-signatures-upl
issue/7-use-case-xades-sign-in-browser
issue/6-use-case-cades-sign-in-browser
issue/12-deploy-self-host-package-conta
issue/3-architecture-doc-autofirma-sig
issue/4-scaffold-web-app-svelte-5-spa-
issue/2-rea-inventory-decompile-autofi
reference/decompile-1.9
Labels
Clear labels   
bullet:blocked

   
bullet:done

   
bullet:queued

   
bullet:review

   
bullet:running
No labels
bullet:blocked
bullet:done
bullet:queued
bullet:review
bullet:running
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
EduarD3V/AutoFirme#6
No description provided.