Skip to content

Feat/vscode session allowlist#1

Merged
Goki602 merged 6 commits intomainfrom
feat/vscode-session-allowlist
Mar 25, 2026
Merged

Feat/vscode session allowlist#1
Goki602 merged 6 commits intomainfrom
feat/vscode-session-allowlist

Conversation

@Goki602
Copy link
Copy Markdown
Owner

@Goki602 Goki602 commented Mar 25, 2026

No description provided.

Goki602 and others added 6 commits March 5, 2026 08:05
…llback

VSCode doesn't support the "ask" permission, so confirm actions are
force-denied. This adds a session allowlist so that non-high-risk
operations auto-allow on retry within the same session, with a softer
denial message guiding users to simply re-instruct Claude.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ude dialogs

Reframe ClawGuard from "security tool" to "confirmation fatigue solver":
- Rewrite README/LP messaging: convenience-first, security as side-effect
- Add `claw-guard stats` command showing auto-allow counts
- Expand session allowlist from VSCode-only to all environments
- Return explicit `permissionDecision: "allow"` instead of null so Claude
  Code actually skips its permission dialog for safe commands
- Simplify billing to free-only model (MIT, no license keys)
- Natural Japanese copy for README.ja.md and LP jp.ts

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…nd improve rule coverage

Replace VSCode-specific confirm-to-deny fallback with a universal deny+retry
pattern: all confirm decisions now return deny with an explanation + retry hint,
pre-registering the session allowlist so the retry auto-allows. This eliminates
the need for vsCodeCompat flag and works consistently across all environments.

Security hardening:
- Restrict CORS to localhost origins only
- Add 1MB request body size limit with early abort
- Add anti-evasion hint to deny/ask responses

Robustness:
- Audit reader skips malformed JSONL lines instead of crashing
- Rule loader catches invalid regex patterns gracefully

Rule improvements:
- Broaden npm/pip install regex to catch flag variants (--ignore-scripts, -D, -g, --no-deps, --upgrade)
- Add isHistoricallyAllowed() for cross-session memory lookup
- Extend session allowlist to cover high-risk confirm decisions

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…l auto-allow tests

- Apply biome formatting across all packages
- Add 4 new Phase 1 core rules (env-file-read, npm-install, pip-install, ssh-key-read) with active marketplace status
- Replace non-null assertions with type-safe casts in integration tests
- Suppress noUselessConstructor for FeatureGate (intentional API contract)
- Add historical auto-allow integration tests (medium/high risk thresholds)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…mparison section, llms.txt, keywords)

- Replace "license key" with "API key" in CLI commands
- Remove plan display from `claw-guard test` output
- Add "Why ClawGuard?" comparison table to README (EN + JA)
- Add llms.txt for LLM crawlers
- Expand npm keywords for search discoverability
- Update billing package description to reflect free model

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Allow users to disable telemetry by setting reputation.opt_in: false
in clawguard.yaml, as documented in README.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@Goki602 Goki602 merged commit 19c0503 into main Mar 25, 2026
1 of 3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant