Conciseness

Quotes: "Reproduce on current `main` with `--format json --quiet`. Save one command that shows the wrong result. Change no code before this command exists." — ~30 lines of pure operational directives with zero concept explanations. Even the build note ("A release build adds a slow LTO link and gives the same findings") is non-obvious project knowledge that earns its tokens; it is not 4 because there is no over-explanation left to trim.

Actionability

Quotes: "`--format json --quiet`", "clear the `.fallow/` cache first", "`cargo build -p fallow-cli`", "a regression test that fails without the fix" — concrete flags, commands, and precise pass/fail criteria throughout. Not 5: the reproduce command is a fragment rather than a complete invocation, "Reduce to a minimal fixture" and step 8's "Run `review` with the affected surface reviewers" are directions without exact commands; not 3 because the guidance is largely executable as written.

Workflow Clarity

Quotes: eight numbered steps with explicit validation gates ("Change no code before this command exists", "a regression test that fails without the fix", "compare old versus new output" on a real consumer) and a genuine error-recovery feedback loop ("After 3 failed fixes, stop and write down the premise that they share. Question that premise before a fourth fix."). Not 4: no validation gap is visible — reproduce-first, failing-test gate, and real-consumer comparison are all explicit.

Progressive Disclosure

The skill is under 50 lines, self-contained, and well organized as a single numbered procedure with a heading; no bundle files (references/, scripts/, assets/) exist to split content into. The single external pointer ("test evidence in `docs/development/quality-gates.md`") is one level deep and clearly signaled, so the simple-skill exception applies and there are no organization gaps to justify a 4.

Specificity

Quotes: "Diagnose and fix a Fallow false positive or false negative" plus "extraction, resolution, graph, analysis, reporting, and real-consumer verification" — several concrete actions and named pipeline layers. Not 3 (it goes beyond 1-2 actions to enumerate the full trace path) and not 5 (the enumerated layers are abstract stage nouns, and the actual user-facing actions compress to just diagnose/fix/verify).

Completeness

The "what" is clear ("Diagnose and fix a Fallow false positive or false negative"), but there is no "Use when..." clause or equivalent trigger guidance anywhere in the description. Per the judging guidelines, a missing explicit trigger clause caps completeness at 3; it is not 4 because the 'when' is entirely absent rather than merely implicit-weak.

Trigger Term Quality

Quotes: "Fallow", "false positive", "false negative", "diagnose", "fix" — the canonical natural terms a user of a static-analysis tool would say. Not 5 (no synonyms such as "wrong result", "spurious finding", "missed issue") and not 3 (the core domain keywords users actually utter are present).

Distinctiveness Conflict Risk

Quotes: "Fallow false positive or false negative" — anchored to a specific proprietary tool and specific defect classes, giving a clear niche with distinct triggers and minimal conflict risk with any other skill. Adjacent anchor 4 would require some overlap concern, which the tool-name anchoring eliminates.