| name | canon-inspect-artifacts |
| description | Use when a Canon run already exists and you need the emitted artifact paths rather than run state or evidence lineage. |
Canon Inspect Artifacts
Support State
available-now
default visibility: discoverable-standard
Purpose
Expose artifact paths from a real Canon run without inventing runtime state.
When To Trigger
- The user wants emitted artifact paths for a run.
When It Must Not Trigger
- No run id is available.
- The user is asking for request-level decisions or evidence lineage instead.
Required Inputs
Preflight Profile
- Verify
canon is on PATH. If missing, point to the install guide.
- Verify
.canon/ exists. If missing, point to $canon-init.
- If no run id was provided, resolve the latest run from Canon runtime state.
- Verify the selected run exists before invoking Canon.
- Do not show preflight checks to the user. Report only the specific missing input.
Canon Command Contract
- Canon command:
canon inspect artifacts --run <RUN_ID>
- Return emitted artifact paths from Canon output only.
- Treat only Canon-backed paths under
.canon/artifacts/<RUN_ID>/... as valid run artifacts.
Expected Output Shape
- concise artifact summary
- real run id
- artifact path pointers
- clear callout of the most important artifact to review first
- no run-level TOML pointers in standard user-facing output
- optional action chips that preserve the same run context, with
Approve generation... only after packet review and Inspect evidence when rationale is the real next need
- ordered possible actions from the current run state
- one recommended next step that preserves the same run context
Failure Handling Guidance
- If
.canon/ is missing, point to $canon-init.
- The preflight keeps the retry surface limited to
RUN_ID.
- If the run id is missing or invalid, ask only for the exact run id and show the exact retry form
canon inspect artifacts --run <RUN_ID>.
- If Canon fails after preflight succeeds, report it as a Canon-execution outcome rather than a preflight failure.
- Never summarize artifacts that were not returned by Canon-backed output.
- Never elevate ad-hoc root-level notes or analysis files into the artifact summary.
- If Canon returns an empty artifact list, say that explicitly and do not imply that a packet exists somewhere else.
Next-Step Guidance
- If the artifact packet explains an approval-gated or recommendation-only block, recommend
$canon-approve next only after the user has reviewed it.
- Use
$canon-inspect-evidence when the user needs the policy rationale or request lineage behind the artifact packet.
- Use
$canon-status only to re-check overall run state after inspection, approval, or resume.
- If a host renders chips, keep
Inspect evidence available when rationale is still missing and never render an approval chip without the real Canon target.
Related Skills
$canon-status
$canon-inspect-evidence