Installing kimi-cli 1.38.0 locally (\`uv tool install kimi-cli\`) and running
it once revealed the plugin's session-discovery logic was built on wrong
assumptions about the on-disk layout.
Observed layout (kimi-cli 1.38.0):
~/.kimi/sessions/<md5(cwd)>/<session-uuid>/
context.jsonl — conversation history
wire.jsonl — turn events (TurnBegin/TurnEnd with user_input payload)
Differences from my original assumptions:
- Sessions are nested under \`sessions/\` (not direct subdirectories of
\`~/.kimi/\`).
- The workspace is identified by an MD5 hash of the absolute path, not by
a \`cwd\` field stored in a state file.
- There is no \`state.json\`. No \`title\`, \`model\`, or \`cost\` is persisted.
- The session ID is the UUID directory name and is accepted as-is by
\`kimi --resume <uuid>\`.
- The old \`--continue\` fallback is unnecessary — if we found the directory,
we always know its UUID.
Fixes:
- \`findKimiSessionMatch\` now computes \`md5(workspacePath)\` with node:crypto
and lists \`~/.kimi/sessions/<hash>/\` directly. No more full-tree scan of
\`~/.kimi/\`, no more \`readFile\` of a fictional \`state.json\`.
- \`getKimiLiveSignalMtime\` keeps the parallel \`Promise.all\` stat of
context.jsonl + wire.jsonl (the only files that exist).
- \`getSessionInfo\` streams the first \`TurnBegin\` out of wire.jsonl as a
best-effort summary, with a 1 MB byte ceiling. agentSessionId is the UUID.
- \`getRestoreCommand\` drops the \`--continue\` fallback branch — a found dir
always has a usable UUID.
Verified end-to-end against the real kimi-cli 1.38 binary on this machine:
- \`detect()\` → true
- \`getLaunchCommand\` output parses cleanly when run with \`--help\`
- \`getSessionInfo\` extracts the actual first user prompt ("say hello")
- \`getRestoreCommand\` produces the same UUID kimi itself prints as the
resume hint: \`kimi -r 6ec34626-aedf-4659-a061-c5fbfa4cf166\`
Tests remain at 75 green. Coverage is now against real on-disk layouts
using temp directories with MD5-hashed bucket names — no mock-structure
drift from reality.
All changes are non-behavioral perf/style cleanups flagged during my second
review pass — no user-visible changes.
- Consolidate double JSON.parse in findKimiSessionMatchUncached: the previous
pass parsed each candidate state.json once to extract cwd and a second time
to extract session_id/model/title. Replaced both helpers with a single
`parseKimiState(raw)` that returns all four fields in one traversal.
- Carry state.json's mtime through KimiSessionMatch so getKimiLiveSignalMtime
(renamed from getKimiSessionMtime) doesn't re-stat state.json — the winner's
mtime was already captured during the scan. Live-signal probe is now limited
to context.jsonl + wire.jsonl (the per-turn files) and runs them in parallel
via Promise.all instead of sequential awaits.
- Fold state.json mtime and the live-signal mtime into a single "freshest"
timestamp in getActivityState so a recently-written context.jsonl wins even
when state.json is stale.
- Tighten appendApprovalFlags signature: `string | undefined` → proper
`AgentPermissionInput | undefined` so typos at call sites fail at compile
time.
- Stricter detect(): don't trust every binary named `kimi` — verify the
--version output mentions kimi/kimi-cli/kimi-code, and fall back to
`kimi info` for builds that print a bare version number. Rejects unrelated
tools that happen to install a `kimi` binary.
Tests: 71 → 75. New coverage:
- detect() accepts kimi-cli vendor strings
- detect() falls back to `kimi info` when --version is ambiguous
- detect() rejects an unrelated `kimi` binary
- Native signal picks the fresher of state.json vs context.jsonl mtimes
Critical (from @harshitsinghbhandari, verified against kimi-cli source):
- Remove `promptDelivery: "post-launch"` — `-p`/`--prompt` is just a prompt
string alias (also `--command`/`-c`), NOT a mode switch. The non-interactive
flag is `--print`, which we never set. Inline delivery via `--prompt` is
reliable and avoids the post-launch sendMessage() delay.
- Drop unchecked `as string` casts in getRestoreCommand in favor of typeof
guards + `?? undefined` so null model values don't silently leak.
Medium (performance):
- Add 30s per-workspace cache to findKimiSessionMatch (mirrors codex's
SESSION_FILE_CACHE_TTL_MS) so the ~/.kimi/ scan doesn't run 12×/min per
active session. Cache keyed by workspacePath; cleared via the new
`_resetSessionMatchCache` test-only export between test cases.
Minor (correctness):
- Collapse findKimiSessionDir + readKimiSessionState into one
findKimiSessionMatch that returns {dir, state} from a single state.json
read. Previously the file was parsed twice per getSessionInfo /
getRestoreCommand call.
- Wire config.subagent → `kimi --agent <name>` (default / okabe / custom).
- Tighten detectActivity patterns so "I approve of this approach" and
"Earlier I failed to connect" no longer falsely trigger waiting_input /
blocked. Regexes are now line-anchored with `^`/`$` + `\b` word boundaries.
Tests: 58 → 71 (all green). New cases cover:
- Native-signal ready/idle decay (previously only active was tested)
- Cascade ordering: JSONL waiting_input wins over a matching native signal
- Malformed state.json in both getSessionInfo and getRestoreCommand
- `work_dir` alias accepted in addition to `cwd`
- project.agentConfig.model preferred over state.json's recorded model
- False-positive narration guards for both regex tightenings
The shared gh wrapper now extracts PR URLs with a regex instead of embedding a literal github URL, so the old codex assertion was stale and broke CI on PR #1300.
Represent missing activity probes as first-class signal states so lifecycle inference only treats valid idle evidence as proof. This prevents false stuck transitions, keeps API/UI lifecycle truth aligned, and makes root monorepo verification deterministic by serializing recursive build and typecheck.
Token sources in streamCodexSessionData are precedence-ordered via `continue`.
`total_token_usage` is a cumulative snapshot (overwrite) while the others are
per-turn deltas (accumulate) — document this so a future reader doesn't "fix"
the asymmetry and break cumulative totals.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Use ?? instead of || for ownerRepo fallback (semantically correct for
null-to-undefined conversion)
- Extract requireRepo() result into a local variable in tracker-gitlab's
updateIssue and issueUrl to avoid redundant validation calls
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Make `repo` field optional in ProjectConfig and Zod schema so projects
without a detected GitHub remote can still load and run
- Remove placeholder `repo: "owner/repo"` from autoCreateConfig() and
addProjectToConfig() — omit the field entirely when no remote is found
- Always use actual workingDir for `path` instead of unreliable `~/<projectId>`
fallback for non-git directories
- Add null guards for `project.repo` across SCM plugins, tracker plugins,
lifecycle manager, webhooks, and prompt builders to prevent crashes when
repo is not configured
Closes#1154
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Use a single StringDecoder across reads so multi-byte UTF-8 sequences
that straddle the 8KB chunk boundary buffer correctly instead of
producing U+FFFD replacement characters that break JSON.parse.
Also fix the test mock: makeFakeFileHandle now advances an internal
cursor and returns bytesRead: 0 at EOF. The prior mock copied from
offset 0 every call, which would infinite-loop readJsonlPrefixLines
for any line larger than the 8192-byte chunk size.
Add a regression test using 3,000 CJK characters (9,000 bytes of
payload) to exercise the chunk boundary path.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Real Codex sessions emit records like
`{"type":"event_msg","payload":{"type":"error",...}}` and
`{"type":"event_msg","payload":{"type":"approval_request",...}}`.
readLastJsonlEntry only exposed the top-level `type`, so the codex
plugin's activity switch matched `event_msg` and decayed to ready/idle,
never surfacing `blocked` or `waiting_input`. The approval_request/error
branches were dead code for payload-wrapped sessions, which is the exact
format this PR series is migrating to.
- readLastJsonlEntry now returns payloadType alongside lastType.
- Codex getActivityState prefers payloadType when present and classifies
task_started/agent_reasoning as active, task_complete as ready, and
the approval/error variants as waiting_input/blocked.
- New tests cover the payload-wrapped approval_request, exec_approval_request,
error, task_started, and task_complete cases end-to-end.
- Core utils gains coverage for payloadType extraction and null fallbacks.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* chore: release 0.2.5
Realign main with npm registry after off-branch publish of 0.2.3/0.2.4.
Bump all 21 linked packages to 0.2.5 and cherry-pick the startup-grace-period
fix for #989 (was in 5e4244a8 but never merged to main).
Also sync non-linked plugin versions (notifier-discord, notifier-openclaw,
scm-gitlab, tracker-gitlab) to their current npm versions.
* Revert "chore: release 0.2.5"
This reverts commit eb17f32834.
* chore: bump all package versions to 0.2.5, remove release workflow
- Bump all 25 packages to 0.2.5 to realign with npm registry
- Update package-version test to expect 0.2.5
- Remove stale .changeset/linear-spawn-branch-name.md
- Delete .github/workflows/release.yml (changesets-based NPM publish)
---------
Co-authored-by: Prateek <karnalprateek@gmail.com>
Co-authored-by: AO Bot <ao-bot@composio.dev>
Fix shell injection vulnerability when combining systemPromptFile with
prompt. The prompt could contain shell metacharacters ($(), backticks)
that would be executed inside the double-quoted string.
Now uses the exact same pattern as OpenCode:
"$(cat 'file'; printf '\n\n'; printf %s 'prompt')"
The shellEscape wraps prompt in single quotes (no shell expansion),
and printf %s outputs it literally without interpretation.
Co-Authored-By: Claude <noreply@anthropic.com>
Use $(cat file) shell substitution instead of inlining file content
to avoid tmux truncation for large system prompt files (2000+ chars).
This matches the pattern used by Claude Code, Aider, and OpenCode.
- Replace readFileSync with $(cat) in getLaunchCommand
- Remove unused readFileSync import
- Update tests to verify shell substitution behavior
Co-Authored-By: Claude <noreply@anthropic.com>
1. Add symlink check for .cursor directory in extractCursorSummary
to match getCursorSessionMtime behavior (prevents path traversal)
2. Add vitest alias for @aoagents/ao-plugin-agent-cursor in CLI tests
(fixes missing module resolution in tests)
3. Add lstatSync check before readFileSync in getLaunchCommand
to reject symlinked systemPromptFile paths (security hardening)
4. Add test coverage for symlink rejection behavior
Co-Authored-By: Claude <noreply@anthropic.com>
- Resolve @composio → @aoagents package renaming conflicts
- Add cursor agent to BUILTIN_PLUGINS in plugin-registry.ts
- Add cursor agent to AGENT_PLUGINS in detect-agent.ts
- Add cursor agent import and registration in plugins.ts
- Add cursor agent dependency and import in web services.ts
- Update cursor plugin package naming to @aoagents/ao-plugin-agent-cursor
- Add cursor agent to changeset linked group
- Fix test imports to use new @aoagents package naming
Co-Authored-By: Claude <noreply@anthropic.com>
Fixes all issues identified in PR review from illegalcall:
1. 🔴 detect() false positives - Now checks for multiple Cursor-specific
markers: "Cursor Agent" text OR (--approve-mcps AND --sandbox flags).
Provides redundancy if Cursor changes one indicator.
2. 🔴 systemPromptFile/systemPrompt ignored - Properly reads file content
synchronously using readFileSync and prepends to prompt. Clean approach
without shell command substitution. Gracefully handles missing files.
3. 🟡 Process regex too generic - Fixed regex from /\\.?/ to /\.?/ for
optional dot prefix. Now correctly matches "agent" or ".agent" binaries.
4. 🟡 Idle check before waiting_input - Reordered detectActivity checks so
waiting_input patterns (permission prompts) are tested BEFORE idle prompt
detection. Fixes false negatives when prompts end with input cursor.
5. 🟡 Symlink/path traversal protection - Added lstat() checks in
extractCursorSummary and getCursorSessionMtime to reject symlinks and
verify paths stay under workspacePath.
6. 🟡 hasRecentCommits false actives - Added comment acknowledging the
limitation (same pattern as Aider plugin). Better than missing activity.
7. 🟡 Missing test coverage - Added 11 new tests:
- 6 tests for detect() covering text match, flag fallback, edge cases
- 5 tests for systemPromptFile/systemPrompt handling including errors
Total: 62/62 tests passing
Co-Authored-By: Claude <noreply@anthropic.com>
Activity Detection (2 related issues):
- Fix getCursorSessionMtime to stat .cursor/chat.md file instead of directory
- Directory mtime only updates on entry changes, not file modifications
- Now checks chat.md file first (tracks actual writes), falls back to directory
- Prevents directory mtime from blocking JSONL fallback in getActivityState
- Allows tier 4 (getActivityFallbackState) to run when needed
Prompt Safety:
- Add -- separator before positional prompt argument in getLaunchCommand
- Prevents prompts starting with - from being misinterpreted as flags
- Matches pattern used in Codex agent plugin
- Update test expectations to include -- separator
Process Detection:
- Update comment to accurately describe "agent" binary matching
- Removed misleading reference to "cursor and .cursor" process names
Plugin Detection:
- Improve detect() to check --version output for Cursor-specific text
- Reduces false positive risk from generic "agent" command name
- Validates output contains "cursor" or "agent" keywords
All tests passing (51/51).
Fixes issues identified in PR #637 review.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
The --trust flag only works in headless mode (with --print), so it doesn't
prevent the workspace trust prompt in interactive mode. Changed to --sandbox
disabled which skips workspace trust prompts entirely.
This fixes the issue where Cursor agent would block on startup waiting for
user to approve the workspace trust prompt.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Renames all npm package scopes from @composio/* to @aoagents/* and
updates GitHub repo references from ComposioHQ/agent-orchestrator
to aoagents/ao throughout the codebase.
- All package.json names and dependencies
- README badges, links, and install instructions
- Documentation references
- Changeset config
- Source code imports and test files
* fix(lifecycle): reduce GitHub API rate limiting from batch enrichment bypass
Three optimizations to prevent API storms in the lifecycle manager poll cycle:
1. **CRITICAL - maybeDispatchMergeConflicts**: Gate the getMergeability()
fallback to only run when batch enrichment didn't run at all. Previously
it called getMergeability() (3 REST calls) whenever hasConflicts was
undefined, even when the batch had already fetched PR data. Now uses
cachedData.hasConflicts ?? false when the batch ran.
2. **HIGH - maybeDispatchCIFailureDetails**: Use batch enrichment ciChecks
when available instead of calling getCIChecks() (separate REST call)
on every poll. The GraphQL batch query now fetches statusCheckRollup
contexts (individual check names, statuses, URLs) alongside the rollup
state. Falls back to getCIChecks() only when batch didn't run.
3. **MEDIUM - maybeDispatchReviewBacklog**: Throttle getPendingComments +
getAutomatedComments API calls to at most once per 2 minutes per session.
These were called every 30s even when nothing had changed.
Impact: ~8-10 API calls/PR/poll reduced to ~2-4, enabling 3-4x more
concurrent sessions before hitting GitHub's 5,000/hr REST limit.
Also extends PREnrichmentData with ciChecks?: CICheck[] and adds
parseCheckContexts() helper to graphql-batch.ts for parsing CheckRun
and StatusContext nodes from the GraphQL statusCheckRollup.contexts field.
* fix(scm-github): fall back to getCIChecks() when contexts list is truncated
When a PR has >20 CI checks, contexts(first: 20) silently truncates the
list. Setting ciChecks to undefined when pageInfo.hasNextPage is true
ensures maybeDispatchCIFailureDetails falls back to the getCIChecks()
REST call, which returns all checks without truncation.
Also adds pageInfo { hasNextPage } to the contexts GraphQL query so
truncation can be detected.
* fix(lifecycle): prune lastReviewBacklogCheckAt in pollAll cleanup loop
Add the new throttle map to the existing pruning loop that removes stale
entries for sessions no longer in the session list. Previously the map
was only cleared on terminal status transitions, leaving orphaned entries
for sessions removed externally (killed + cleaned up without transition).
* fix(lifecycle): bypass throttle on review transition; fix StatusContext conclusion
Two fixes for automated review findings:
1. Bypass review backlog throttle when a transition reaction just fired for
humanReactionKey or automatedReactionKey. The transitionReaction branch
needs to read the current fingerprint via the API to record
lastPendingReviewDispatchHash. Without bypassing, the throttle prevents
this write and the next unthrottled poll sees a stale (empty) hash,
clears the reaction tracker, and fires a duplicate dispatch.
2. Set conclusion on StatusContext nodes in parseCheckContexts() to match
the REST getCIChecksFromStatusRollup() format (rawState.toUpperCase()).
The CI failure fingerprint includes c.conclusion ?? '', so inconsistent
conclusion values between GraphQL and REST paths caused phantom fingerprint
changes when switching sources, triggering duplicate dispatches.
* fix(scm-github): normalize CheckRun conclusion and map NEUTRAL to skipped
Two consistency fixes in parseCheckContexts() vs the REST path:
1. NEUTRAL conclusion: was mapped to 'passed' (with SUCCESS), but
mapRawCheckStateToStatus() in the REST path maps NEUTRAL to 'skipped'.
Changed to treat NEUTRAL the same as SKIPPED.
2. CheckRun conclusion: was stored as the raw GraphQL string (may be
lowercase). REST getCIChecks/getCIChecksFromStatusRollup always store
conclusion as rawState.toUpperCase(). Now stores rawConclusion which
is already uppercased during the status branching logic.
Both fixes prevent phantom fingerprint changes when maybeDispatchCIFailureDetails
switches between GraphQL batch and REST fallback across poll cycles.
* fix(scm-github): map STALE/NOT_REQUIRED/NONE conclusions to skipped
parseCheckContexts() was mapping these conclusions to 'failed' via the
else fallback, while mapRawCheckStateToStatus() in the REST path
explicitly maps all of them to 'skipped'. Added them to the skipped
branch alongside SKIPPED and NEUTRAL to fully mirror the REST mapping.
* fix(scm-github): map QUEUED/WAITING to pending not running
parseCheckContexts() mapped QUEUED and WAITING CheckRun statuses to
'running', but mapRawCheckStateToStatus() in the REST path maps both
to 'pending'. Only IN_PROGRESS maps to 'running' in the REST path.
Fixes fingerprint inconsistency when switching between GraphQL batch
and REST fallback across poll cycles.
* fix(scm-github): map STARTUP_FAILURE to skipped; guard null pageInfo
- STARTUP_FAILURE conclusion now falls through to the "skipped" branch
(matching mapRawCheckStateToStatus() REST default) instead of the
explicit failure enumeration catch-all
- Null pageInfo guard prevents TypeError from typeof null === "object"
JavaScript quirk when accessing hasNextPage on a null pageInfo field
- Tests added for both cases
* fix(scm-github): map COMPLETED+null conclusion to skipped not passed
When a CheckRun has status COMPLETED and conclusion null, the REST path's
mapRawCheckStateToStatus() converts it to "" which maps to "skipped".
The GraphQL path was incorrectly mapping it to "passed" via !rawConclusion.
Fix: only map rawConclusion === "SUCCESS" to "passed"; null falls through
to the else branch → "skipped", matching the REST path exactly.
* fix(agent-claude-code): return idle state when no JSONL session file exists
Freshly spawned sessions had no Claude Code JSONL file yet (Claude Code
doesn't create it until the first conversation), causing getActivityState
to return null → displayed as 'unknown' in ao status.
When the process is running but no session file exists, return
{ state: 'idle', timestamp: now } so the dashboard shows the correct
state immediately after spawn.
Closes#883
* fix(agent-claude-code): use session.createdAt for idle timestamp when no JSONL file
Using new Date() as the timestamp caused isIdleBeyondThreshold to always
compute ~0ms, preventing stuck detection from ever firing for sessions that
hang before creating a JSONL file. Using session.createdAt correctly
represents when the session began, allowing stuck detection to eventually
trigger.
Harsh Batheja
harshitsinghbhandari
Harshit Singh Bhandari
Dhruv Sharma
Adil Shaikh
adil
yyovil
i-trytoohard
Gautam Tayal
Prateek
Harsh Batheja
Denis Darii