fix(security): stop leaking internal error details from web API routes (CWE-209)

[groupthinking]

Canonical issue

Closes #

Outcome

Next.js API routes no longer return raw exception strings (String(error), error.message, details: String(...)) to clients. Each handler now responds with a static, non-revealing message and logs the full error server-side via console.error. This removes a reconnaissance vector where backend hostnames, connection diagnostics, or stack context could leak to callers.

Scope

• Included:
  • apps/web/src/app/api/agents/dispatch/route.ts, agents/status/route.ts — drop details: String(error) from 502 responses.
  • apps/web/src/app/api/extract-events/route.ts — replace leaked error.message with "Event extraction failed".
  • apps/web/src/app/api/jobs/[jobId]/route.ts — add server-side console.error, return generic "Failed to fetch job status".
  • apps/web/src/app/api/training/status/route.ts, training/trigger/route.ts — drop details: String(...); add console.error for the upload/trigger failure path so the detailed error is preserved internally without reaching the client.
  • apps/web/src/app/api/video/search/route.ts — replace leaked error.message with a generic message.
  • .jules/sentinel.md — records the vulnerability/learning/prevention note.
• Explicitly excluded: backend (Python) error handlers; no changes to response status codes or success payloads.

Risk

• Risk level: low
• Failure mode: A client that previously parsed the leaked details/message string for diagnostics would now see only a generic message. No internal control flow depends on these strings.
• Rollback: revert the branch; changes are isolated to catch-block response bodies.

Verification

List exact automated and manual checks, tied to the current head SHA.

• Focused tests
• Static review: confirmed no dangling references to removed bindings (message in extract-events, uploadError in training/trigger) and no remaining unused catch bindings.
• Required CI
• Review threads resolved

Note: the JS/TS toolchain (eslint/tsc) is not installed in the automation sandbox, so lint/typecheck was performed by static inspection only. CI should run the authoritative checks.

Production evidence

Not applicable at PR-open time — awaiting CI / Vercel preview on the current head.

Agent handoff

• One canonical issue is linked
• No competing PR implements the same issue
• Acceptance criteria are satisfied
• Required checks pass on the current head
• Human decision is requested only for product, security, irreversible infrastructure, or production approval — this PR is left as draft pending human merge approval to the protected main branch.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NFDfJ3C6qTJ2Xda8SW93Pf

---

Generated by Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
v0-uvai	Ready	Preview, v0	Aug 1, 2026 6:37pm

[coderabbitai]

Important

Review skipped

Auto reviews are limited based on label configuration.

🏷️ Required labels (at least one) (1)

• [‘architecture-gap’, ‘bug’, ‘ci-cd’, ‘ci/cd’, ‘copilot-rabbit’, ‘documentation’, ‘duplicate’, ‘enhancement’, ‘frontend’, ‘github_actions’, ‘good first issue’, ‘help wanted’, ‘high-priority’, ‘invalid’, ‘javascript’, ‘ml-model’, ‘needs-triage’, ‘pipeline-critical’, ‘placeholder-code’, ‘priority:high’, ‘python’, ‘python:uv’, ‘question’, ‘styling’, ‘tests’, ‘v0’]

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository YAML (base), Repository UI (inherited), Organization UI (inherited)

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 5dcb0913-e0f7-4b2c-9704-1b7f79f0dcf7

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Snapshot Warnings

⚠️: No snapshots were found for the head SHA e5e91a3.

Ensure that dependencies are being submitted on PR branches and consider enabling retry-on-snapshot-warnings. See the documentation for more information and troubleshooting advice.

Scanned Files

None

[github-actions]

Agent Completion Truth Gate: BLOCKED

Reasons: invalid_payload

Machine-readable verdict

{
  "details": {
    "invalid_fields": [
      "issue.number",
      "policy.agent_login",
      "policy.run_id"
    ]
  },
  "reasons": [
    "invalid_payload"
  ],
  "verdict": "blocked"
}

Workflow evidence

[groupthinking]

Automated remediation status — HALTED(awaiting_human_governance)

The code change itself is sound: Vercel build deployed green (Ready), and validate / label / Canonical issue and evidence checks passed. I also fixed a lint/security regression the original commits introduced (an unused uploadError catch binding that had dropped internal logging — now restored via console.error). Labels added (security, javascript, frontend) so CodeRabbit's auto-review can run.

Two checks are red, and neither is a defect in this diff — both need a human decision, so I'm not force-passing them:

1. gitleaks (working tree) — pre-existing false positive, not caused by this PR.
The finding is uv.lock:5129, rule square-access-token, on:

hash = "sha256:eaaac4c9fdd5e9e8852dc778d2d7405897ec510f2a298071453e5e3a07914bb1"  # parso 0.8.7 sdist

That is a PyPI package integrity hash, not a secret. This PR does not touch uv.lock, so the gate is red on main too — it's a repo-wide .gitleaks.toml allowlist gap. Suggested fix (separate change, since editing secret-scan config doesn't belong in an error-leakage PR): add uv.lock to the [allowlist] paths (lockfiles carry only package hashes), or allowlist the square-access-token rule for *.lock files.

2. agent-completion/truth-gate + Agent completion enforcement — governance gate, needs a human.
blocked: invalid_payload, invalid fields issue.number, policy.agent_login, policy.run_id. This PR has no linked canonical issue and no trusted agent-provenance manifest. This gate exists precisely to require human-governed provenance, so I won't fabricate a run_id/agent_login to slip past it.

To move this forward, a maintainer needs to: link a canonical tracking issue (fills the Closes # / agent-lock manifest), and clear the uv.lock gitleaks false-positive repo-wide. The PR is intentionally left as draft pending merge approval to the protected main branch.

---

Generated by Claude Code

[groupthinking]

Closing as a duplicate of #1145

This PR turned out to duplicate #1145 — both carry the identical 8-file error-leakage fix on the same head commit f856d46. #1145 is the canonical version: it's authored by Jules, is ready-for-review (not draft), and links the owning task 1121709532317009963. This branch happened to be sitting on the same two Sentinel commits, so opening #1147 created a competing PR for the same change.

Per the repo's "no competing PR implements the same issue" rule, I'm closing this in favor of #1145. The only delta this PR added — a console.error('Training upload/trigger error:', uploadError) in the training/trigger catch block (the branch #1145 modified but left without internal logging) — is carried over as a suggested follow-up comment on #1145.

---

Generated by Claude Code