Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is highly actionable with concrete tools, parameters, and policies, and its workflow is clearly sequenced with explicit gating and checklists. Its main weaknesses are minor redundancy in the evidence/instruction-promotion guidance and the absence of any progressive file structure for the specialized modes.
Suggestions
Merge the two adjacent sentences about evidence-to-instruction promotion into a single rule to remove the redundancy and tighten token use.
Consider splitting the Code Review Memory and TaskFlow Work Log modes into separate reference files (e.g. references/code-review.md) referenced from the main body, so the general workflow stays lean and the specialized detail is progressively disclosed.
Add a one-line worked example of a recall or write-back call with concrete parameter values to make the already-actionable guidance copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean and operational with no padding about concepts Claude already knows, but it could be tightened — notably the two adjacent sentences about memory promotion ('Agent-written memory starts as evidence by default...') and ('Decision memories can become future instructions only when... user_confirmed or imported...') restate the same evidence-to-instruction rule, matching 'mostly efficient but includes some unnecessary explanation or could be tightened'. | 2 / 3 |
Actionability | For an instruction-only skill the guidance is concrete and specific: exact tool names (openbrain_recall, openbrain_writeback, etc.), explicit parameter lists (task_type, query, entities, scope, limits, sensitivity), enumerated write-back categories, and named use_policy fields; per the scoring notes, absence of code is not penalized when guidance is this actionable. | 3 / 3 |
Workflow Clarity | The Core Rule states the sequence (recall before work, write back after), sections follow that order, and there are explicit validation/gating steps (use_policy levels, project_only/include_unconfirmed defaults), a feedback loop (conflict-resolution rule), a fallback (tools unavailable -> continue and note), and checklist-style mode sections, matching 'clear sequence with explicit validation steps; feedback loops; checklists'. | 3 / 3 |
Progressive Disclosure | The skill is a single ~110-line file with well-organized sections and no nested/deep references (good), but it exceeds the 'under 50 lines' allowance for a no-reference skill and keeps two specialized modes (Code Review Memory, TaskFlow Work Log) inline that could be externalized, matching 'some structure but content that should be separate is inline'. | 2 / 3 |
Total | 10 / 12 Passed |