Add SafeSkill security badge (88/100 — Passes with Notes) - #1
Open
OyaAIProd wants to merge 1 commit into
Open
Conversation
Signed-off-by: SafeSkill Scanner <mk@oya.ai>
spuentesp
added a commit
that referenced
this pull request
Aug 19, 2026
P0 #1 from the post-plan gap analysis: 'auth/rate-limit' (gate producto). Previously the HTTP transport was open to any IP that reached the port — no auth, no rate limit. An attacker could register_agent with any name, claim files, and manipulate the state. An LLM in a tight loop could saturate the server. This commit adds: - src/server/auth.rs: new module with AuthState (Bearer API key + token-bucket rate limit) and BearerToken + RateLimit helpers. Five unit tests cover the auth + rate-limit contract. - src/server/ingest/mod.rs: LainServer gains an 'auth: Arc<AuthState>' field, populated from env at server startup (LAIN_API_KEYS, LAIN_RATE_LIMIT_RPM, LAIN_RATE_LIMIT=off). - src/server/mcp/handler.rs: the HTTP request path now does an auth + rate-limit check before any other routing. /health is exempt (operational probe). 401 on missing/invalid bearer; 429 with Retry-After header on rate-limit exceeded. Stdio transport bypasses this entirely. - tests/auth_integration.rs: 8 runtime integration tests (bearer parsing, dev mode, multi-key acceptance, rate-limit per-key bucketing, refill timing, disable semantics). - tests/federation_integration.rs: pre-existing test expected the pre-040cdd6 behavior (BeyondCurrent → empty changed_symbols). Updated to match the post-040dd6 contract (Retracted entries preserved across the error path). Env vars: LAIN_API_KEYS=key1,key2,key3 comma-separated LAIN_RATE_LIMIT_RPM=60 default; set 0 to disable LAIN_RATE_LIMIT=off explicit disable If LAIN_API_KEYS is unset, AuthState is in dev mode: every request passes auth (no rate limit either if not configured). Stdio transport (Claude Code, Cursor, etc.) is exempt from both since the agent process is local. Verification: - 13/13 smoke6 assertions against a live HTTP server in 4 scenarios: dev mode (no env), auth on + /health exempt, per-key rate-limit bucketing, LAIN_RATE_LIMIT=off override. - 8/8 auth_integration integration tests in cargo test. - 568 lib + 16 federation_integration (after 1 updated) + 8 auth_integration + all other binaries 0 failures. Total tools: still 64 (no new MCP surface; auth is a transport gate, not a tool). The product is now safe to expose on a network without rate-limit-vulnerability to a misbehaving LLM loop.
spuentesp
added a commit
that referenced
this pull request
Aug 19, 2026
Resolves the gap analysis ask '_meta.revision en todas las read tools' (P1 #1). Previously only the overlay's monotonic counter was exposed on the JSON-RPC envelope. Agents using read tools could not tell how fresh the static graph was without a separate list_repos round-trip. This commit adds alongside on every envelope (stdio and HTTP paths): - The stdio path's calls with where the value is — Unix epoch seconds of the most recent successful re-index start, or when the re-index is skipped / failed / timed out. - The HTTP path's closure builds the envelope inline (its own literal) and was also updated to inject the same field. Both paths read the value once per dispatch so it is stable for the duration of the response. - The value is also cached in tests/static_graph_meta.rs as 5 integration tests covering skipped / ok / failed / timeout outcomes, plus a 'fresh server starts skipped' sanity check. Verification: - 5/5 integration tests pass. - 568 lib + 16 federation + 8 auth + 32 + others = 0 failures. - Live HTTP server: GET list_repos returns {_meta: {revision: 0, static_graph_generation: null}} on fresh boot and {revision: N, static_graph_generation: <ts>} after request_reload.
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.
Top Findings
npm-shim/bin/lain.js:6)npm-shim/bin/lain.js:20)npm-shim/bin/lain.js:6)npm-shim/bin/lain.js:20)npm-shim/bin/lain.js:7)View full report on SafeSkill
About SafeSkill
SafeSkill is a free, open-source security scanner for AI tools, MCP servers, and Claude Code skills. We scan for code exploits, prompt injection, and data exfiltration risks.
False positive? We take accuracy seriously. If any finding above is incorrect, please open an issue and we will fix it immediately.