Security: Redact sensitive values from command output#21
Open
fredluz wants to merge 1 commit intogarrytan:mainfrom
Open
Security: Redact sensitive values from command output#21fredluz wants to merge 1 commit intogarrytan:mainfrom
fredluz wants to merge 1 commit intogarrytan:mainfrom
Conversation
Prevent secrets and sensitive data from leaking into stdout and model transcripts by redacting output from type, cookie, header, forms, and storage commands. Fixes garrytan#18 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
garrytan
added a commit
that referenced
this pull request
Mar 13, 2026
type no longer echoes text (reports character count), cookie redacts value with ****, header redacts Authorization/Cookie/X-API-Key/X-Auth-Token, storage set drops value, forms redacts password fields. Prevents secrets from persisting in LLM transcripts. 7 new tests. Credit: fredluz (PR #21) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
garrytan
added a commit
that referenced
this pull request
Mar 13, 2026
* Phase 2: Enhanced browser — dialog handling, upload, state checks, snapshots - CircularBuffer O(1) ring buffer for console/network/dialog (was O(n) array+shift) - Async buffer flush with Bun.write() (was appendFileSync) - Dialog auto-accept/dismiss with buffer + prompt text support - File upload command (upload <sel> <file...>) - Element state checks (is visible/hidden/enabled/disabled/checked/editable/focused) - Annotated screenshots with ref labels overlaid (-a flag) - Snapshot diffing against previous snapshot (-D flag) - Cursor-interactive element scan for non-ARIA clickables (-C flag) - Snapshot scoping depth limit (-d N flag) - Health check with page.evaluate + 2s timeout - Playwright error wrapping — actionable messages for AI agents - Fix useragent — context recreation preserves cookies/storage/URLs - wait --networkidle / --load / --domcontentloaded flags - console --errors filter (error + warning only) - cookie-import <json-file> with auto-fill domain from page URL - 166 integration tests (was ~63) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * Phase 2: Rewrite SKILL.md as QA playbook + command reference Reorient SKILL.md files from raw command reference to QA-first playbook with 10 workflow patterns (test user flows, verify deployments, dogfood features, responsive layouts, file upload, forms, dialogs, compare pages). Compact command reference tables at the bottom. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * Phase 3: /qa skill — systematic QA testing with health scores New /qa skill for systematic web app QA testing. Three modes: - full: 5-10 documented issues with screenshots and repro steps - quick: 30-second smoke test with health score - regression: compare against saved baseline Includes issue taxonomy (7 categories, 4 severity levels), structured report template, health score rubric (weighted across 7 categories), framework detection guidance (Next.js, Rails, WordPress, SPA). Also adds browse/bin/find-browse (DRY binary discovery using git rev-parse), .gstack/ to .gitignore, and updated TODO roadmap. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * Bump to v0.3.0 — Phase 2 + Phase 3 changelog Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * feat: cookie-import-browser — Chromium cookie decryption module + tests Pure logic module for reading and decrypting cookies from macOS Chromium browsers (Comet, Chrome, Arc, Brave, Edge). Supports v10 AES-128-CBC encryption with macOS Keychain access, PBKDF2 key derivation, and per-browser key caching. 18 unit tests with encrypted cookie fixtures. * feat: cookie picker web UI + route handler Two-panel dark-theme picker served from the browse server. Left panel shows source browser domains with search and import buttons. Right panel shows imported domains with trash buttons. No cookie values exposed. 6 API endpoints, importedDomains Set tracking, inline clearCookies. * feat: wire cookie-import-browser into browse server Add cookie-picker route dispatch (no auth, localhost-only), add cookie-import-browser to WRITE_COMMANDS and CHAIN_WRITE, add serverPort property to BrowserManager, add write command with two modes (picker UI vs --domain direct import), update CLI help text. * chore: /setup-browser-cookies skill + docs (Phase 3.5) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * chore: bump version and changelog (v0.3.1) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * security: redact sensitive values from command output (PR #21) type no longer echoes text (reports character count), cookie redacts value with ****, header redacts Authorization/Cookie/X-API-Key/X-Auth-Token, storage set drops value, forms redacts password fields. Prevents secrets from persisting in LLM transcripts. 7 new tests. Credit: fredluz (PR #21) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * security: path traversal prevention for screenshot/pdf/eval (PR #26) Add validateOutputPath() for screenshot/pdf/responsive (restricts to /tmp and cwd) and validateReadPath() for eval (blocks .. sequences and absolute paths outside safe dirs). 7 new tests. Credit: Jah-yee (PR #26) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: auto-install Playwright Chromium in setup (PR #22) Setup now verifies Playwright can launch Chromium, and auto-installs it via `bunx playwright install chromium` if missing. Exits non-zero if build or Chromium launch fails. Credit: AkbarDevop (PR #22) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * security: fix path validation bypass, CORS restriction, cookie-import path check - startsWith('/tmp') matched '/tmpevil' — now requires trailing slash - CORS Access-Control-Allow-Origin changed from * to http://127.0.0.1:<port> - cookie-import now validates file paths (was missing validateReadPath) - 3 new tests for prefix collision and cookie-import path traversal Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: address review informational issues + add regression tests - Add cookie-import to CHAIN_WRITE set for chain command routing - Add path validation to snapshot -a -o output path - Fix package.json version to match 0.3.1 - Use crypto.randomUUID() for temp DB paths (unpredictable filenames) - Add regression tests for chain cookie-import and snapshot path validation Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * docs: add /qa, /setup-browser-cookies to README + update BROWSER.md - Add /qa and /setup-browser-cookies to skills table, install/update/uninstall blurbs - Add dedicated README sections for both new skills with usage examples - Update demo workflow to show cookie import → QA → browse flow - Update BROWSER.md: cookie import commands, new source files, test count (203) - Update skill count from 6 to 8 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * feat: team-aware /retro v2.0 — per-person praise and growth opportunities - Identify current user via git config, orient narrative as "you" vs teammates - Add per-author metrics: commits, LOC, focus areas, commit type mix, sessions - New "Your Week" section with personal deep-dive for whoever runs the command - New "Team Breakdown" with per-person praise and growth opportunities - Track AI-assisted commits via Co-Authored-By trailers - Personal + team shipping streaks - Tone: praise like a 1:1, growth like investment advice, never compare negatively Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * docs: add Conductor parallel sessions section to README Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
typecommand: No longer echoes typed text; reports character count instead (Typed 12 characters)cookiecommand: Redacts cookie value in output (Cookie set: name=****)headercommand: Redacts values for sensitive headers (Authorization, Cookie, Set-Cookie, X-API-Key, X-Auth-Token); non-sensitive headers remain visibleformscommand: Replaces password field values with[redacted]instead of dumping actual valuesstorage setcommand: Drops the value from the confirmation messageWhy
In an LLM-agent context, command output is fed back into model context and persisted in logs/transcripts. Commands like
type,cookie, andheaderwere reflecting secrets (passwords, auth tokens, API keys) directly into stdout, creating a security risk.Fixes #18
Test plan
browse type <password>and verify output saysTyped N characterswithout echoing the textbrowse cookie session=secret123and verify output showsCookie set: session=****browse header Authorization:Bearer token123and verify output showsHeader set: Authorization: ****browse header Content-Type:application/jsonand verify non-sensitive header value is still shownbrowse formsand verify password values show[redacted]browse storage set key valueand verify output showsSet localStorage["key"]without the value🤖 Generated with Claude Code