openclaw-android-node-apk / scripts
AGENTS.md59 starsChanged 16 days agoMITView on GitHub
Own local tooling, script wrappers, generated-artifact helper rules. Repo-global policy: root AGENTS.md; read only task-relevant references.
- Installs packages
We split this AGENTS.md into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 23
Never bare vitest ... in automation: starts local watch mode unless run or --run explicit.
- Very specific: line 25
Before checks, inspect changed-file scope: node scripts/check-changed.mjs --dry-run [--staged|-- <files...>].
- Very specific: line 37
Before PR: narrow proof — touched tests (pnpm test <file> --maxWorkers=1) + targeted typecheck/lint/format.
- Very specific: line 46
Before returning: join all PR-state-mutating children.
- Distinctive: line 49
Before comments/ownership-checked cleanup: confirm merge receipt.
≈ “Common” is an estimate: at least 1 in 5 of the 36,368 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Scripts Guide
Own local tooling, script wrappers, generated-artifact helper rules. Repo-global policy: root AGENTS.md; read only task-relevant references.
Ownership Map
- Tests:
run-vitest.mjs; follow testing skill + test-authoring skill. - Lint/typecheck:
run-oxlint.mjs+run-tsgo.mjsown repo-local runtime behavior. - Changed-file checks:
check-changed.mjsowns execution;changed-lanes.mjsowns classification. Never copy path-scope rules into hooks/CI snippets. - PR preparation:
pr/pr-libown source acquisition, correction reviews, gate bindings. Changing them: read PR preparation. - PR wrapper lifecycle:
pr/pr-libown supervision, locks, materialized dependencies. Changing/repairing them: read PR wrapper tooling. - GitHub transport:
pr-libowns observations + protected dispatch. Changing API/Octopool behavior: read PR GitHub transport. - Merge admission/recovery:
pr-libowns policy, retained outcomes, completion. Read merge admission, merge recovery, or prior-CI admin admission for that operation. - Remote proof:
crabbox-wrapper.mjsowns source capsules;.github/actions/prepare-testbox-shellowns shell/cwd. Changing them: read remote gate contracts. Operating boxes: use Crabbox skill. - Generated artifacts: align source generator, package script, verification command. Prefer
*:gen/*:checkpairs over undocumented one-offs.
Wrapper Rules
- macOS-only Bash scripts:
#!/bin/bash, Bash 3.2-compatible syntax. Invoke directly or with/bin/bash, including fixtures. - Portable Bash entrypoints using heredocs/here-strings (including sourced helpers): inline Darwin Bash 5.3+ re-exec guard before those operations prevents heredoc pipe deadlocks. stdin/sourced installers: request
/bin/bashwhen replay impossible. - Existing curated repo seam: prefer wrappers over raw tool entrypoints.
- Tests: prefer
scripts/run-vitest.mjsor rootpnpm test ...entrypoints over rawvitest runcalls. - Never bare
vitest ...in automation: starts local watch mode unlessrunor--runexplicit. - Adding/editing package scripts or CI steps honoring repo-local lint/typecheck runtime behavior: prefer
scripts/run-oxlint.mjs+scripts/run-tsgo.mjs. - Before checks, inspect changed-file scope:
node scripts/check-changed.mjs --dry-run [--staged|-- <files...>]. - Lint (one/few files): prefer direct
node scripts/run-oxlint.mjs --tsconfig <matching config> <files...>over shardedpnpm lint.check-changed.mjsowns targeting for core, extension, script diffs.
TypeScript Syntax
scripts/**TypeScript implementation files: erasable by Node without transformation. No parameter properties, runtime enums/namespaces, import-equals, export-assignment, or other transform-required TypeScript syntax.- Syntax rule ≠ every script a plain-Node entrypoint. Keep
tsxfor closures intentionally depending on source runtime trees, frozen checkouts, package aliases, or tsconfig/path resolution. - Native Node execution: opt-in per entrypoint + import closure. Only when runtime imports stay Node-resolvable; never pull broader source trees into this syntax policy.
Proof Rules
- PR CI = default broad proof. Change builds + focused tests pass → open ready PR; never hold for pre-PR suite runs.
- Before PR: narrow proof — touched tests (
pnpm test <file> --maxWorkers=1) + targeted typecheck/lint/format. Prefer local execution for trusted, bounded work when dependencies and current host capacity are qualified; local execution is optional. - Boxes: deliberate remote proof for expensive checks, a busy/unqualified local host, another OS/device, live providers/channels, long E2E or heavy benchmarks. Acquire when the command is ready; use the smallest proven profile. Queued → stop only task-owned allocation, use PR CI or a safe local scope; keep platform-specific gaps pending. Never force heavy local checks or chain provider fallbacks.
- CI fails →
pnpm test:ci-failed <pr>; read published classification (pre-existing on main vs new). Fix new failures →pnpm test:ci-failed <pr> --rununtil local pass → push once. Pre-existing: failure-table landing rule. Never rerun/re-push just for green (gh run rerunincluded); never local full suite for CI repair. - Untrusted contributor/fork code: secretless isolation. Credentialed execution needs maintainer authorization; scope live credentials + isolate Gateway/state.
PR Safety
- PR operations: follow maintainer workflow. Unset
OPENCLAW_PR_GATES_REMOTE= GitHub default; explicitOPENCLAW_PR_GATES_REMOTE=githubequivalent. Default handoff:merge-run --auto-merge. Old full local gates: explicitOPENCLAW_PR_GATES_REMOTE=local.OPENCLAW_TESTBOX=1+ unset remote selector retains hosted mode. Pending admission ≠ successful proof/merge completion. - Landing subcommands require canonical/origin-main wrapper code. Never bypass wrapper trust, operation locks, exact-head review/gate bindings, live authority checks, or Octopool string rewrite protection; never select raw
ghto get landing through. - Before returning: join all PR-state-mutating children. Drained self-exited failures auto-release their operation lock; interrupted/uncertain locks stay held — verify no child tools remain before reported exact-OID
scripts/pr lock-recovercommand. Never delete lock refs manually. - Preserve transition journals, preparation evidence, outcome refs, captures. Never clear/push
refs/openclaw/pr-merge-outcomes/<PR>, hand-edit receipts, or infer non-execution from failed command, elapsed time, absent process, or open PR. - Accepted/pending/uncertain writes: investigate through native recovery. Never blindly resubmit/switch transports. Existing land authority covers investigated same-scope recovery, not gate bypasses/new scope/method.
- Before comments/ownership-checked cleanup: confirm merge receipt. Completion comments: one attempt; reconcile uncertain comments by marker. Preserve worktrees/captures with unresolved outcome; never delete recreated branch by name.
Execution Gotchas
Commands apply within task authority/safety boundaries. Prefer bounded local proof when dependencies and current host capacity are qualified.
- Fs-safe dependency trouble: follow on-demand vendoring instructions; vendor contents local, registry dependencies default.
- Trusted normal checkout missing dependencies:
pnpm install --frozen-lockfile, retry once before diagnosing code defect. Never reconcile shared/worktree install while other jobs use it. - CLI:
pnpm openclaw ...orpnpm dev, nevernode --import tsx src/index.ts; supported wrappers own build freshness + process setup. - Formatting: installed
oxfmt; typechecking: repositorytsgolanes. Scope:pnpm changed:lanes --json; targeted tests/checks. Avoiding worktree reconciliation:node scripts/check-changed.mjsornode scripts/run-vitest.mjswith ready dependencies.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 28,693 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository107 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- extensions/AGENTS.mdAGENTS.md
- test/AGENTS.mdAGENTS.md
- ui/AGENTS.mdAGENTS.md
- agent-transcriptSkill
- auto-qaSkill
- autoreviewSkill
- channel-message-flowsSkill
- clawdtributorSkill
- claw-scoreSkill
- clawsweeperSkill
- control-ui-e2eSkill
- crabboxSkill
- discord-clawdSkill
- discord-e2eSkill
- discord-user-postSkill
- discrawlSkill
- gitcrawlSkill
- graincrawlSkill
- notcrawlSkill
- openclaw-changelog-updateSkill
- openclaw-ci-limitsSkill
- openclaw-debuggingSkill
- openclaw-docker-e2e-authoringSkill
- openclaw-ghsa-maintainerSkill
- openclaw-live-updaterSkill
- openclaw-parallels-smokeSkill
- openclaw-pr-maintainerSkill
- openclaw-qa-testingSkill
- openclaw-refactor-docsSkill
- openclaw-release-validationSkill
- openclaw-repair-sweepSkill
- openclaw-secret-scanning-maintainerSkill
- openclaw-test-heap-leaksSkill
- openclaw-testingSkill
- openclaw-test-performanceSkill
- openclaw-updateSkill
- parallels-discord-roundtripSkill
- plugin-sdk-impactSkill
- proof-videoSkill
- prototype-openclaw-tuiSkill
- pr-simplifySkill
- release-openclaw-announcementSkill
- release-openclaw-ciSkill
- release-openclaw-macSkill
- release-openclaw-maintainerSkill
- release-openclaw-plugin-testingSkill
- security-triageSkill
- slack-e2eSkill
- slacrawlSkill
- solidSkill
- tag-duplicate-prs-issuesSkill
- technical-documentationSkill
- telegram-e2e-userbotSkill
- test-auditSkill
- update-team-serverSkill
- verify-releaseSkill
- 1passwordSkill
- apple-notesSkill
- The first 60 shown.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

