autospec-monitor

autospec-monitor

Use this skill when the operator asks for live observability into an autospec-autonomous run: what conductor is running, whether it was launched in real mode or --dry-run, which cycle/tier/action is current, how fresh the heartbeat is, what stop flag is set, and what discovery artifacts were produced.

Autospec monitor is read-only by default. It must not mutate repository state or GitHub state unless the operator explicitly asks it to file a monitor-related issue.

Operating Rules

• Prefer the typed Rust command surfaces over raw log parsing.
• Use autospec autonomous monitor --once for a concise operator snapshot.
• Use autospec autonomous status --json and autospec autonomous timeline --lines N for follow-up detail; do not hand-parse logs unless those typed fields are missing.
• Always distinguish launch mode from tier result:
  • run_mode=dry-run means the conductor launch argv included --dry-run.
  • tier_dry_result.dry=true means that tier filed no candidates in that cycle; it does not by itself mean the conductor was launched with --dry-run.
• Accept and pass through --repo OWNER/REPO and --repo-dir DIR using the same conventions as autospec-autonomous / autospec autonomous.
• Stay useful when fields are partially empty by reporting -, null, or an empty artifact list rather than inventing state.
• GitHub issue writes require explicit operator confirmation and must be limited to monitor-related defects or enhancement requests.

Command Shape

autospec autonomous monitor --once --repo OWNER/REPO --repo-dir DIR

For machine-readable output:

autospec autonomous monitor --once --json --repo OWNER/REPO --repo-dir DIR

For a broader timeline after the snapshot:

autospec autonomous status --json --repo OWNER/REPO --repo-dir DIR
autospec autonomous timeline --lines 80 --repo OWNER/REPO --repo-dir DIR

What To Report

Summarize these fields when present:

• conductor identity: repo, PID, log path, launch argv, stop flag path/mode;
• launch mode: real versus dry-run from launch argv;
• current work: cycle, tier, action, state status, and heartbeat age;
• tier dry result: explain dry=true as a no-filed-candidates tier result, not as proof of launch --dry-run mode;
• discovery artifacts: .autospec/explore-once-* metadata, proposal counts, candidate/filed issue titles, and verification mode.

Intent Mapping

Operator asks	Do this
What is autospec-autonomous doing?	Run autospec autonomous monitor --once --repo OWNER/REPO --repo-dir DIR.
Is this a real run or dry-run?	Report run_mode from monitor output and quote the launch argv evidence.
Did Tier 2/4 dry mean dry-run?	Explain that tier dry=true means no filed candidates for that tier cycle.
What did discovery find?	Report .autospec/explore-once-* proposal/candidate counts and titles.
Fields are missing or empty.	Fall back to status/timeline/artifact summaries and mark unknown fields as - or null.
File an issue about monitor output.	Confirm the requested write, then file only that monitor-related issue.