Manage CI Artifacts Skill

Persistence layer for CI run artifacts. The ci-verify finalize step
calls this skill to write a per-job log slice plus a manifest.toon
under artifacts/ci-runs/{run_id}/ inside the plan directory at
classification
time. The eager-fetch model keeps the evidence on disk before
retrospectives run, immune to GitHub's 90-day log retention window.

Enforcement

Base contract: See manage-contract.md for shared enforcement rules, TOON output format, and error response patterns.

Skill-specific constraints:

• --run-id MUST be a non-empty string; the value is the per-run directory
  key and an empty value would collide every run into the same directory.
• The persist subcommand is idempotent: a second invocation for the same
  (plan_id, run_id) MUST be a no-op that re-emits the existing manifest
  (no log re-fetching).
• All CI log fetching MUST flow through the
  plan-marshall:tools-integration-ci:ci fetch-logs abstraction; no
  direct gh / glab calls inside this skill's scripts.

Storage Location

.plan/plans/{plan_id}/
  artifacts/
    ci-runs/
      {run_id}/
        manifest.toon      # one per run — enumerates every job
        {job_name}.log     # per-job log slice
        ...
      {next_run_id}/       # loop-back commit produces a new run dir
        manifest.toon
        ...

The manifest.toon shape is documented in
`standards/persistence-layout.md`.
Multi-run loop-back cascades MUST NOT overwrite previous runs — every
{run_id} directory is independent and survives until the plan dir is
archived.

---

Operations

Script: plan-marshall:manage-ci-artifacts:manage-ci-artifacts

persist

Fetch and write the full run artifacts (eager mode).

python3 .plan/execute-script.py plan-marshall:manage-ci-artifacts:manage-ci-artifacts \
  persist --plan-id {plan_id} --run-id {run_id} --head-sha {sha} \
  --pr-number {pr_number} --provider {github|gitlab}

Output (TOON):

status: success
plan_id: {plan_id}
run_id: {run_id}
run_dir: artifacts/ci-runs/{run_id}
already_persisted: true | false
job_count: N
manifest_path: artifacts/ci-runs/{run_id}/manifest.toon
log_paths[N]:
  - artifacts/ci-runs/{run_id}/{job_name}.log
  - ...

Idempotence: when artifacts/ci-runs/{run_id}/manifest.toon already
exists, the script returns already_persisted: true and re-emits the
existing manifest contents (paths and job_count) without re-fetching logs.

read

Read a previously persisted manifest. Used by the retrospective phase.

python3 .plan/execute-script.py plan-marshall:manage-ci-artifacts:manage-ci-artifacts \
  read --plan-id {plan_id} --run-id {run_id}

Output (TOON): The persisted manifest content.

list

Enumerate all persisted runs under the plan dir, sorted by fetched_at.

python3 .plan/execute-script.py plan-marshall:manage-ci-artifacts:manage-ci-artifacts \
  list --plan-id {plan_id}

Output (TOON):

status: success
plan_id: {plan_id}
run_count: N
runs[N]{run_id,head_sha,fetched_at,job_count}:
  ...

---

Retention

Default: keep all runs for the lifetime of the plan dir. The plan dir is
itself the unit of retention — when archive-plan archives the plan, the
artifacts/ci-runs/ tree archives with it. No automatic pruning at
phase-6-finalize completion; opt-in pruning is out of scope for this
skill and would be a future enhancement gated by user demand. See
`standards/persistence-layout.md` for
the no-prune decision rationale.

Canonical invocations

The canonical argparse surface for manage-ci-artifacts.py. The plugin-doctor analyzer (_analyze_manage_invocation.py) reads this section as source-of-truth for the manage-invocation-invalid and missing-canonical-block rules. Consuming docs xref this section by name instead of restating the command inline. See `pm-plugin-development:plugin-script-architecture` cross-skill-integration.md § "Script invocation in documentation".

persist

python3 .plan/execute-script.py plan-marshall:manage-ci-artifacts:manage-ci-artifacts persist \
  --plan-id PLAN_ID --run-id RUN_ID --head-sha HEAD_SHA --pr-number PR_NUMBER \
  --provider {github,gitlab} \
  [--jobs-file JOBS_FILE] [--wait-outcome {completed,deadline_exceeded}] [--final-status FINAL_STATUS]

read

python3 .plan/execute-script.py plan-marshall:manage-ci-artifacts:manage-ci-artifacts read \
  --plan-id PLAN_ID (--run-id RUN_ID | --latest)

--run-id and --latest are mutually exclusive; exactly one must be supplied.

list

python3 .plan/execute-script.py plan-marshall:manage-ci-artifacts:manage-ci-artifacts list \
  --plan-id PLAN_ID