Content
51%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a maximally lean generated stub that delegates entirely to an external shared skill file. While nothing is wasted token-wise, it contains no actionable workflow of its own, no safety/validation guidance for the destructive "discard" option, and a reference that does not resolve within the skill bundle.
Suggestions
Inline the core workflow (or bundle it in a references/ file): include the merge/PR/keep/discard decision options and their steps in SKILL.md or a bundled file, rather than relying solely on the external `~/.agents/skills/branch-finish/SKILL.md` path, which is not part of this skill bundle and may not resolve at runtime.
Add explicit validation/safety guidance for the destructive "discard" option — e.g., confirm with the user before discarding and check for unmerged/uncommitted work — since missing feedback loops for destructive operations cap workflow clarity.
Resolve the stale reference: the comment cites `scripts/agent-harnesses.py`, but no scripts/ directory exists in the bundle — either bundle the generator script or drop the reference so the file listing matches the cited paths.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is two lines — "Load and follow the shared skill at `~/.agents/skills/branch-finish/SKILL.md`" plus a one-line provenance comment — with zero padding and zero explanation of concepts Claude already knows; every token is functional, matching anchor 5. The cost of this leanness shows up in the other dimensions, not here. | 5 / 5 |
Actionability | The only guidance is the pointer "Load and follow the shared skill at `~/.agents/skills/branch-finish/SKILL.md`" — a specific, executable directive in itself, but every step for the actual task (presenting merge/PR/keep/discard options) is absent from the body, matching anchor 2's minimal-concrete-guidance profile. Not anchor 1 because the pointer is concrete rather than vague; not anchor 3 because no partial task guidance exists. | 2 / 5 |
Workflow Clarity | No sequence for the skill's actual workflow appears in the body; the single coherent delegation instruction (load, then follow) is a rough gappy sequence at best, and there is no validation guidance for the destructive "discard" option — matching anchor 2. Anchor 1 would require incoherence, and the written instruction is unambiguous; anchor 3 would require listed steps, which are absent. | 2 / 5 |
Progressive Disclosure | The one reference is clearly signaled and exactly one level deep, but scored against the actual bundle structure: the referenced file lives outside the bundle at an environment-dependent path (`~/.agents/...`), no references/scripts/assets directories exist, and the comment cites `scripts/agent-harnesses.py` which is not bundled — structure exists but the package is incomplete, sitting between anchors 2 and 4. | 3 / 5 |
Total | 12 / 20 Passed |