Content
60%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 operationally rich with executable code, explicit validation guards, and well-signaled one-level-deep references. Its dominant weakness is verbosity — heavy philosophical rationale and repeated meta-principles inflate the core file well beyond an overview.
Suggestions
Move the duplicated guard/enforcement philosophy into a single reference (e.g. observation-log.md) and reference it once, instead of restating 'the empty result is a claim about the instrument' across the scan, id-derivation, and archival sections.
Pull the long 'Observed:' anecdotes out of the operational prose into a short rationale reference or footnotes, keeping only the rule each anecdote motivates inline.
Trim the dense prose inside each Session Start Protocol step into a crisp checklist line first, then the rationale, so the sequence reads as a navigable workflow rather than embedded essays.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~800-line body is noticeably verbose: the same guard/enforcement philosophy is restated across multiple sections ('the write is the enforcement', the empty-result guard rule), interspersed with long 'Observed:' anecdotes and parenthetical asides that could be tightened without losing the operational signal. | 2 / 5 |
Actionability | Provides fully executable bash snippets (find/awk/sed/mv with noclobber guards), a concrete YAML frontmatter template, and a Quick Reference table — mostly copy-paste-ready guidance with only minor abstract prose gaps such as 'Set that point before the second hypothesis'. | 4 / 5 |
Workflow Clarity | The Session Start Protocol is a clear numbered sequence (1–7) with explicit validation guards ('SCAN COMMAND BROKEN... exit 1', 'ID COMMAND BROKEN', collision/noclobber checks) and feedback loops (re-derive id on collision), but the checkpoints are buried in dense prose rather than crisp standalone steps, keeping it just below 5. | 4 / 5 |
Progressive Disclosure | A dedicated 'Reference files — load on demand' section signals seven real one-level-deep reference files, each named with its load trigger; however the core SKILL.md itself retains substantial inlined operational detail (snippets, file format, archival logic) that keeps it from a clean 5. | 4 / 5 |
Total | 14 / 20 Passed |