agentsclimarketplace
Sign in

Event logging

Skill bartoszarendt/agenticloop/skills/event-logging

Use when a role needs to emit an Agentic Loop workflow-gate event (role.invoked, task.created, task.started, check.run, review.started, review.result, decision.recorded, blocked, needs_context, task.closed, summary.published) and must resolve the event logging command, write concise event entries, or confirm that event logging is disabled or unavailable. Owns the command-resolution recipe, the disabled/non-blocking rules, and the event data conventions.From its SKILL.md

Install
npx -y skills add bartoszarendt/agenticloop --skill event-logging

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.1 KB, as published. Nobody here has run it

Event logging

Event logging is optional and disabled by default. It writes local workflow-gate events to .agenticloop/logs/<TASK-ID>.jsonl through the Node CLI. agenticloop/AGENTIC_LOOP.md owns why event logging exists, the event taxonomy, and which lifecycle gates emit which events. This skill owns the operational procedure: resolving the command, honoring the disabled and non-blocking rules, and keeping entries concise.

When event logging is off

Read .agenticloop/project.md before writing any event. Agents must not attempt CLI event logging unless it says event_logging: enabled. When event logging is disabled, do nothing: writing events is a no-op, and skipping them is not a process gap.

Resolving the command

When event_logging: enabled, resolve the event logging command once per host session before the first event write:

  1. If event_logging_command is non-empty, use that command.
  2. If event_logging_command is blank or omitted, run npx agenticloop --help once. If it succeeds, use npx agenticloop.
  3. If no working event logging command is available, do not repeatedly retry and do not block the workflow. Record a truthful process gap in the task record, review, or closeout marker note, then continue.

After resolution, event writes use:

<resolved-command> event-logging <event_type> --task <TASK-ID> --role <role> --summary "<short fact>"

agenticloop event remains a compatibility alias, but new instructions should use event-logging. Default writes target .agenticloop/logs/<TASK-ID>.jsonl via --task <TASK-ID>; use --output <file> only for tests or an explicit local exception. Add --outcome only for event types that require it.

Unavailable logging is always non-blocking: never fail, pause, or loop a task because event logging could not run.

Required and recommended fields

Every event needs --task <TASK-ID> (when a task is known), --role, and a short --summary. Events in the same target that share a task_id derive the same deterministic trace_id unless --trace-id is supplied. Use refs for identifiers (github:issue:<n>, github:pr:<n>, commit:<sha>, branch:<name>, task-file:<path>, command:<command>) and --data-json for small structured context.

For a fresh Auditor delegation, the Orchestrator records role.invoked with data.target_role: "auditor", data.invocation_reference, data.delegation_mode, and data.fallback: false. When logging is enabled, audit report cross-checks that event; event logging remains optional and an unavailable logger never manufactures verified provenance.

When event logging is enabled, a completed or reviewed task ending with zero required gate events is non-conformant. Do not backfill missed events as if they happened on time; record the miss as a process gap instead.

Keep entries small

Entries are summaries, not transcripts. Do not write raw prompts, raw assistant text, full tool output, transcript payloads, token streams, per-turn telemetry, or host runtime exports. Keep data small, structured, and non-transcript. Command evidence belongs in the durable task artifact, not the event log; the event carries concise verdict facts and references only.

Command safety

When an event summary or data value contains Markdown or shell-significant characters, avoid inlining it through fragile shell quoting. Prefer the safe payload mechanism named by the backend or delegation prompt, keep scratch under .agenticloop/tmp/, and use relative forward-slash paths.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.