fix(security): neutralize CR/LF in rendered log records (CWE-117 formatter gap)

[groupthinking]

Canonical issue

Related to (not implementing) #913 / #898. #913 is the canonical issue for #810's router-level _safe_log work; this PR is complementary and closes the formatter-level CWE-117 gap Copilot flagged on #810, which lives outside router.py. It does not fulfil #913's file contract, so it should not be scored against it. This work needs its own human-chartered canonical issue (see Agent provenance).

Outcome

StructuredFormatter returned the base-formatted string verbatim, so any CR/LF carried by user-controlled data or by raw exception text appended via exc_info / logger.exception could forge or split log lines. Inline message sanitization (e.g. the _safe_log work on #810's router) never reaches the exc_info traceback or structured extra fields, leaving a log-injection vector on the fully rendered record.

This escapes CR/LF and other line separators in the single central sink — StructuredFormatter.format() — so every logger call (message, extra, and exc_info traceback) is neutralized at once, even when a call site forgets to sanitize its inputs. Escaping (not dropping) keeps the original content visible and greppable while preventing it from starting a new physical line.

Scope

Single concern — two files only:

• src/youtube_extension/backend/config/logging_config.py — add _UNSAFE_LOG_CHARS table; format() now .translate()s the final rendered record.
• tests/unit/test_logging_config_crlf.py — new regression suite asserting rendered handler output (message, exc_info, logger.exception+extra, and full control-char coverage).

Explicitly excluded: router.py message interpolation (owned by #810); no shared logsafe module is introduced (none exists on main — the fix is self-contained).

Note: an earlier revision of this branch also carried three code_generator.py scaffold-cleanup commits. Those are the canonical content of PR #1252 (closes #1253), so they were rebased out of this branch (head is now 4ed812e) — no work was discarded, and this PR is now a focused single-concern security change.

Risk

• Risk level: low
• Failure mode: multi-line tracebacks now render as a single line with escaped \n separators — a deliberate CWE-117-safe trade-off (one physical line per record). Ordinary single-line messages are byte-for-byte unchanged (test_ordinary_message_is_unchanged).
• Rollback: revert the two-line format() change; no schema, API, or data migration involved.

Verification

Tied to head 4ed812ee4aea4b79ea26604b9e21adb6fa58ac37:

• Focused tests — tests/unit/test_logging_config_crlf.py: 15 passed (PYTHONPATH=src pytest ... -o addopts="").
• Before/after proof against rendered output: stock formatter → 5 physical lines, forged CRITICAL - FORGED ADMIN LINE on its own line; patched formatter → 1 physical line, payload inline and escaped.
• ruff check clean on both changed files.
• agent-completion/truth-gate — BLOCKED, human-gated (see below).

Production evidence

Backend Python change; alters only the textual representation of log records, not request values or behavior. The Next.js apps/web Vercel preview does not exercise this Python path.

Agent handoff

• Single-concern scope; does not compete with fix(security): sanitize user-controlled values in API logs (CWE-117 log injection) #810/Execution: sanitize v1 router log inputs (CWE-117) (PR #810) #913, and no longer duplicates fix(codegen): clean up orphaned scaffold on cancelled/failed generation #1252's codegen commits
• Acceptance criteria satisfied (rendered-output regression tests green)
• agent-completion/truth-gate green — cannot be satisfied autonomously
• Human decision requested for: chartering a canonical issue with a frozen pre-dispatch intent snapshot, then un-drafting + final security review + merge to protected main

Agent provenance

This PR is intentionally not accompanied by a fabricated agent-lock-manifest. The truth-gate requires a canonical issue with a frozen pre-dispatch intent snapshot and a trusted terminal agent-result; those are pre-dispatch governance artifacts a post-hoc autonomous run cannot honestly manufacture without impersonating the gate. Per the guidance on #810 ("do not weaken or impersonate the gate"), the manifest is left for a human to charter — hence the gate's invalid_payload. The code-level evidence above stands on its own.

---

🤖 Generated with Claude Code

https://claude.ai/code/session_01BQcuj14cL5gGjdy2Rb925v

[github-actions]

🔍 PR Validation

✅ Current validation passed.

[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: 76bbd432-1b4b-4605-b43a-d23778156c37

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.}

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
v0-uvai	Ready	Preview, v0	Aug 2, 2026 5:58pm

[github-actions]

Dependency Review

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

Snapshot Warnings

⚠️: No snapshots were found for the head SHA 4ed812e.

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": [
      "policy.agent_login",
      "policy.run_id"
    ]
  },
  "reasons": [
    "invalid_payload"
  ],
  "verdict": "blocked"
}

Workflow evidence

[groupthinking]

@coderabbitai review

Truth-gate disposition (unattended remediation run)

The agent-completion/truth-gate is red, and I am not turning it green autonomously — this is the correct, honest terminal state. Breakdown against the machine verdict:

Code-level checks are green on head 9fb4a8a:

• tests/unit/test_logging_config_crlf.py — 15 passed (rendered-output regression: message, exc_info traceback, logger.exception+extra, full control-char set).
• ruff check clean; Dependency Review ✅; Vercel preview Ready.

What I corrected this run:

• Removed the mis-linked agent-lock-manifest that claimed Execution: sanitize v1 router log inputs (CWE-117) (PR #810) #913 — Execution: sanitize v1 router log inputs (CWE-117) (PR #810) #913 is fix(security): sanitize user-controlled values in API logs (CWE-117 log injection) #810's canonical router issue, so scoring this formatter fix against it produced the scope_drift / missing_declared_files / agent_run_id_mismatch noise. This PR is complementary to fix(security): sanitize user-controlled values in API logs (CWE-117 log injection) #810/Execution: sanitize v1 router log inputs (CWE-117) (PR #810) #913, not a competing implementation.
• Added accurate labels (copilot-rabbit, security, python, tests) so Copilot/CodeRabbit review is no longer skipped.

What remains blocked — and requires a human, by design:

1. missing_intent_snapshot — the gate needs a frozen pre-dispatch intent snapshot. I cannot manufacture one after the fact without impersonating the gate (per the explicit warning on fix(security): sanitize user-controlled values in API logs (CWE-117 log injection) #810).
2. scope_drift (residual) — this branch also carries three pre-existing, unmerged code_generator.py scaffold-cleanup commits. Branch policy forbids discarding unmerged work, so an autonomous run cannot both keep them and match a single-issue file contract. A maintainer should decide whether to split them into their own PR.
3. draft_pr + final human review — this is a security change to protected main; it stays draft pending sign-off. Un-drafting solely to pass the gate would be gaming it.

Staged next command for a maintainer, once chartered:

gh pr ready 1255 && gh pr merge 1255 --squash   # after canonical-issue charter + security review

---

Generated by Claude Code

[coderabbitai]

@groupthinking I will review the formatter change and the regression coverage. I will assess code behavior independently from the human-gated agent-completion/truth-gate status.

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.