agent-command-run
Skill6 starsChanged 3 months agoMITView on GitHub
Run a noisy command and report compact output without losing failure evidence or confusing process outcomes. Use when command output is too large to review interactively. RU: длинный вывод команды, сжать вывод.
In this repository65 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/agent-hook-verify/references/claude.mdCLAUDE.md
- agent-context-auditSkill
- agent-hook-verifySkill
- agent-integration-auditSkill
- agent-model-findSkill
- agent-project-syncSkill
- agent-skill-findSkill
- agent-skill-syncSkill
- agent-terms-defineSkill
- agent-workspace-migrateSkill
- api-contract-defineSkill
- api-docs-syncSkill
- api-stream-diagnoseSkill
- api-yandex-oauth-runSkill
- code-async-reviewSkill
- code-cli-buildSkill
- code-config-auditSkill
- code-durable-reviewSkill
- code-performance-fixSkill
- code-state-verifySkill
- deploy-artifact-publishSkill
- deploy-rollout-runSkill
- design-brand-buildSkill
- docs-claims-verifySkill
- docs-diagram-buildSkill
- docs-prose-fixSkill
- git-branch-cleanSkill
- git-merge-syncSkill
- git-path-runSkill
- git-repo-runSkill
- git-worktree-guardSkill
- infra-container-setupSkill
- infra-network-diagnoseSkill
- infra-project-migrateSkill
- infra-runbook-buildSkill
- kanban-board-auditSkill
- kanban-card-buildSkill
- kanban-card-planSkill
- kanban-card-reviewSkill
- kanban-card-runSkill
- kanban-card-verifySkill
- kanban-epic-coordinateSkill
- kanban-queue-coordinateSkill
- kanban-roadmap-planSkill
- kanban-schedule-runSkill
- monitor-telemetry-setupSkill
- personal-action-planSkill
- quality-artifact-verifySkill
- quality-change-reviewSkill
- quality-change-verifySkill
- research-prior-art-findSkill
- research-source-auditSkill
- research-source-findSkill
- research-spike-runSkill
- security-access-diagnoseSkill
- security-api-hardenSkill
- security-boundary-hardenSkill
- security-boundary-reviewSkill
- security-filesystem-hardenSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 25
When implementing or reviewing structured capture pipelines, use receipt verification for producer, reader and parser boundaries.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
Compact command reporting
Use when command output is too large for useful interactive review. Run the actual authorized command with its arguments unchanged; capture full output to an access-controlled location when needed and permitted. Print a bounded digest with both line and byte limits, a clear truncation indicator and the path to retained evidence without exposing secrets. A short routine command does not need a capture wrapper.
Preserve each command's real exit status, signal, timeout and completion state separately from capture, transport or parser errors. A launched process or partial tail is not a completed result. Check the full retained artifact for relevant failures and counts before reporting success; a search with zero matches is meaningful only if its scope and instrument were validated. Report filtered or truncated excerpts as excerpts.
Capture wrapper
The bundled capture wrapper runs one command and keeps its merged stdout and stderr in a private log (a 0700 directory under TMPDIR, file mode 0600). Resolve the skill location through the current runtime's plugin interface, then run python3 <skill-dir>/scripts/hermes-log-run.py [--tail-lines N] [--tail-bytes N] -- <command and exact argv>. It prints LOG_PATH=<path> first, then one RESULT line, then a bounded tail.
- The
RESULTline carriesclassification(OK,FAIL,INTERRUPTEDorLOG_ERROR), the wrapper and child exit codes,log_stateand the raw size and line counts.classification=OKmeans the child exited0, the log was finalized and no signal interrupted the wrapper; a printed success message alone is never enough. The child's own status (including126and127) is preserved; a logging failure exits125. - The tail defaults to 80 lines and 8192 bytes. Explicit limits must be positive and are capped at 10,000 lines and 65,536 bytes, so a single enormous line cannot flood the context.
- On
INTorTERMthe wrapper forwards the signal to the command's own process group, waits 2 seconds, then sendsSIGKILLto that group. An interrupted run leaves a partial log (log_state=partial); report it as partial, never as a complete result. - The wrapper captures output only. It does not authorize the wrapped command, redact secrets (bounds are not redaction) or clean up old logs; retention is a separate, explicitly confirmed operation.
Self-test: python3 <skill-dir>/scripts/test_hermes_log_run.py --artifact-root <private-dir> --report-path <new-file> uses scripts/fixture_emitter.py as a deterministic child process and needs no network. --report-path must not exist yet.
For structured, mixed or long-running output and retention boundaries, use outcome and capture checks. Do not expose raw credentials or private payloads in logs or excerpts; use the runtime's available interface and follow the project's retention policy.
When implementing or reviewing structured capture pipelines, use receipt verification for producer, reader and parser boundaries.
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.

