Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body delivers a highly actionable, well-sequenced archive workflow with strong validation and feedback loops for a destructive operation; its main weaknesses are minor Guardrails duplication and an all-inline structure with no progressive disclosure.
Suggestions
Remove Guardrails entries that restate steps already covered (change-selection prompting, inline sync, never-archive-while-syncing) to tighten the body and lift conciseness.
Consider extracting the detailed step-4 sync-verification criteria (ADDED/MODIFIED/REMOVED/RENAMED checks) into a referenced reference file so the SKILL.md overview stays lean.
If the sync verification detail is kept inline, tighten its prose to the essential pass/fail criteria to reduce token cost.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient for a complex workflow, but the Guardrails section restates step 1 ('Always prompt for change selection') and step 4 ('If sync is requested, run the openspec-sync-specs workflow inline', 'Never archive while a spec sync is still in flight'), fitting anchor 2's 'could be tightened' rather than the fully lean anchor 3. | 2 / 3 |
Actionability | Concrete executable commands ('openspec list --json', 'openspec status --change "<name>" --json', 'mkdir -p "<planningHome.changesDir>/archive"', 'mv "<changeRoot>" ...'), named JSON fields, and explicit routing logic make it copy-paste ready, matching anchor 3. | 3 / 3 |
Workflow Clarity | Steps 1-6 are clearly sequenced with explicit validation checkpoints (artifact completion, task counts, sync state) and a feedback loop ('If the sync failed, or any capability does not match, report what differs and stop — do not archive') plus a pre-move existence check on a destructive mv, matching anchor 3. | 3 / 3 |
Progressive Disclosure | With no bundle files present the body is well-sectioned (Store selection, Input, Steps, Output, Guardrails), but all content is inline and the lengthy step-4 sync-verification prose is content that could arguably be split out, fitting anchor 2 rather than the 'content appropriately split' of anchor 3. | 2 / 3 |
Total | 10 / 12 Passed |