Content
85%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.
An exceptionally lean body with an explicit amend-until-verdict feedback loop, loud validation checkpoints, and concrete decision rules for deriving the contract. The main defects are the dangling `cheat-museum.md` reference at the judge step and prose compressed hard enough in places (e.g. the draft bullet) that a worked example would help.
Suggestions
Ship the referenced `cheat-museum.md` false-pass catalog (e.g. under references/) or remove the reference — the judge step depends on it and the file is absent from the bundle.
Add one short worked example of a locked contract (observable, reading command, direction, terminal value, cap, invariants) to ground the densely compressed draft bullet.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is 23 lines of prose with zero padding and no explanation of concepts Claude already knows — every sentence carries instruction ("Success is defined before the run, not by it."). Lean and efficient; every token earns its place. | 5 / 5 |
Actionability | Concrete decision rules ("the literal command that reads it (the strongest the work allows: drive the running app for a UI, a live terminal for a TUI, tests otherwise)"), named skills to invoke per condition, and enumerated contract parts make the guidance mostly executable. Minor gaps: no worked example of a finished contract, and the judge step's "known false-pass catalog in `cheat-museum.md`" is not present in the bundle. | 4 / 5 |
Workflow Clarity | Clear condition-driven sequence (clarify → elicit → draft → anchor → judge) with an explicit feedback loop ("Each open false pass judge finds re-enters as an amendment — back through the drafting and anchoring conditions above — until the verdict stands"), explicit validation checkpoints ("A command that will not run fails here, loudly"; "It locks only when judge's verdict stands and the user locks it"), and error recovery for post-run gaming ("amend it here and rerun from the last honest point"). | 5 / 5 |
Progressive Disclosure | Sections are well-organized and the body is short, but its single external reference — "Start from the known false-pass catalog in `cheat-museum.md`" — points to a file that does not exist in the skill directory (no references/ or bundle files are present). A dangling dependency on a core step prevents a higher score despite good structure. | 3 / 5 |
Total | 17 / 20 Passed |