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 has a well-intentioned structure with clear workflow steps, explicit failure modes, and good reference architecture, but suffers significantly from verbosity and repetition. The same safety constraints and philosophical points are restated across multiple sections (Philosophy, Constraints, Execution Boundaries, Safety Boundaries, Anti-Patterns), inflating token cost without adding clarity. The actionability would benefit greatly from a concrete output artifact template showing the expected markdown structure.
Suggestions
Consolidate overlapping sections (Constraints, Execution Boundaries, Safety Boundaries, Anti-Patterns) into a single 'Constraints & Boundaries' section to eliminate redundancy and reduce token cost by ~40%.
Add a concrete, copy-paste-ready output artifact template showing the expected `.harness/linear/**-linear-plan.md` structure with all required fields (schema_version, selected_stage, linear_mutation_status, etc.) filled with example values.
Remove the Philosophy section and the repeated 'Linear is execution state; .harness is cognition' mantra — state it once in the opening line and trust Claude to retain it.
Add intermediate validation checkpoints within the 12-step procedure (e.g., after step 4's Linear state verification and after step 6's filtering) rather than concentrating all validation at step 12.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is extremely verbose (~200+ lines) with extensive domain jargon, repeated concepts, and sections that restate the same constraints in slightly different ways (e.g., 'Execution Boundaries' vs 'Constraints' vs 'Safety Boundaries' overlap significantly). Much of the content describes organizational process that Claude could infer from a much shorter specification. Concepts like 'cognition vs execution state' are repeated multiple times. | 1 / 3 |
Actionability | The procedure provides a numbered 12-step workflow with specific actions, and the examples section gives concrete scenarios with expected behaviors. However, there is no executable code beyond a single script path, no concrete payload templates or schemas showing what the output artifact should look like, and many steps rely on references that aren't provided. The guidance is specific to the domain but lacks copy-paste-ready artifact examples. | 2 / 3 |
Workflow Clarity | The 12-step procedure is clearly sequenced and includes validation at step 12 with fail-fast semantics and explicit gate outcomes. However, the validation checkpoints are mostly at the end rather than interspersed, and the feedback loop for error recovery is mentioned but not clearly structured (e.g., step 12 says 'stop at first failed gate' but doesn't specify recovery paths). The mutation confirmation gate at step 11 is a good checkpoint, but overall the workflow has gaps in intermediate validation. | 2 / 3 |
Progressive Disclosure | The References section provides extensive one-level-deep links to supporting documents, which is good progressive disclosure structure. However, without bundle files to verify these references exist, and given that the main body itself contains substantial inline content that could be offloaded (e.g., the detailed output field requirements, anti-patterns, gotchas), the balance between overview and detail is not well-calibrated. The skill tries to be both an overview and a comprehensive reference simultaneously. | 2 / 3 |
Total | 7 / 12 Passed |