Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill suffers severely from redundancy and verbosity — the same 12 principles are stated three times (raw directive, decoded principles, MUST rules), consuming enormous token budget without adding new information each time. The actionable operational content (workflow detection table, gh commands, validation checklist, agent spawning protocol) is solid but buried under layers of philosophical preamble. The skill would benefit enormously from collapsing the triple-stated principles into a single concise MUST list and moving the owner's raw directive to a separate reference file.
Suggestions
Collapse the three redundant sections (raw owner directive, decoded principles, MUST rules) into a single concise MUST rules list — the current triple-statement wastes ~2000 tokens repeating the same points.
Move the raw owner directive monologue to a separate file (e.g., owner-directive.md) and reference it, keeping only the operational MUST rules inline in SKILL.md.
Add concrete inline examples for at least one complete workflow (e.g., 'triage issue #42') showing the exact sequence of agent spawns, tool calls, and validation steps rather than deferring everything to external files.
Remove narrative backstory and motivation paragraphs (e.g., 'Over the lifetime of context-mode we have shipped at least three high-impact regressions...') — Claude doesn't need the historical context to follow the operational rules.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Extremely verbose — the preamble alone is ~2500 words of raw owner monologue followed by a near-complete restatement as 'decoded principles' and then ANOTHER restatement as 'MUST rules'. The same anti-hallucination, TDD, parallel-agent, and platform-equality points are repeated 3-4 times each. Much of the content explains motivations and backstory ('the owner has been burned...') that Claude doesn't need. The refs/ table and auto-recovery protocol, while useful, are buried in narrative. Token budget is severely disrespected. | 1 / 3 |
Actionability | There are some concrete, actionable elements — the gh CLI commands, the validation checklist, the workflow detection table, and the agent spawning protocol steps are useful. However, most of the skill is philosophical directives and principles rather than executable steps. The actual 'how to triage an issue' or 'how to do a release' is deferred entirely to referenced files (triage-issue.md, release.md, etc.) which are not provided. The git archaeology commands (git log --follow, git log -S) are concrete but appear only once amid walls of directive text. | 2 / 3 |
Workflow Clarity | The high-level workflow (Analyze → Recruit → Dispatch → Ping-pong → Ship) is present and the workflow detection table is clear. The three blocking gates (Claim Verification, TDD-First, Grill-Me) provide validation checkpoints. However, the actual detailed workflows are all delegated to external files (triage-issue.md, review-pr.md, release.md) which aren't provided, so the skill itself doesn't contain complete step-by-step sequences with explicit validation/feedback loops for any concrete workflow. | 2 / 3 |
Progressive Disclosure | The skill does reference external files (agent-teams.md, validation.md, tdd.md, communication.md, marketing.md, triage-issue.md, review-pr.md, release.md) which is good progressive disclosure structure. However, the SKILL.md itself is a monolithic wall of text — the massive preamble, decoded principles, and MUST rules sections contain enormous amounts of inline content that is triply redundant. The overview-to-detail ratio is inverted: the 'overview' file is the longest document. No bundle files were provided to verify the references exist. | 2 / 3 |
Total | 7 / 12 Passed |