Content
31%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 suffers from heavy duplication and boilerplate, and — most seriously — its Quick Start and CLI examples reference an API and flags that the packaged script does not implement, so the main executable guidance fails as written. The workflow and error-handling sections are serviceable but generic, and most listed reference files do not exist in the bundle.
Suggestions
Rewrite the Quick Start and Command Line Usage sections against the actual script surface: e.g. `python scripts/main.py --source physics --target biology --mode bridge --output json --save matches.md`, and remove the nonexistent `scripts.interdisciplinary` / `CollaborationFinder` API (or add that module).
Deduplicate the body: collapse the two "When to Use" sections, two "References" sections, and the verbatim description pastes in "When to Use" and "Key Features", and merge the doubled Dependencies list into one consistent set of pinned versions.
Fix progressive disclosure by removing references to nonexistent files (`references/guide.md`, `references/examples/`, `references/api-docs/`) and either creating them or replacing them with the real bundle contents (e.g., `references/audit-reference.md`).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is noticeably verbose and padded: the frontmatter description is pasted verbatim twice (in "When to Use" line 13, including a malformed "the task needs Use when identifying..." splice, and again in "Key Features" line 19), there are two "When to Use" sections, two "References" sections, and a Dependencies list that duplicates entries (networkx, numpy, sklearn/scikit-learn listed twice with conflicting version specs from two different requirements files). This matches the 2 anchor for several unnecessary padded sections; it is not a 1 because genuine concrete content (Quick Check commands, Handling Poor Results) is present. | 2 / 5 |
Actionability | The headline executable examples do not run against the packaged script: the Quick Start imports `from scripts.interdisciplinary import CollaborationFinder`, but no such module or class exists in `scripts/` (main.py defines KnowledgeGraph/BridgeAnalyzer/OutputFormatter), and the CLI example uses flags (`--my-field`, `--target-field`, `--find-collaborators`, `--output matches.json`) that the actual argparse surface does not accept (`--source`, `--target`, `--mode`, `--output {json,markdown}`). The only accurate commands are the Quick Check (`python -m py_compile scripts/main.py`, `python scripts/main.py --help`), leaving minimal correct concrete guidance — a 2, not a 3, because the primary code/CLI guidance is fabricated relative to the real script rather than merely incomplete. | 2 / 5 |
Workflow Clarity | A sequenced 5-step Workflow with a validation gate ("Validate that the request matches the documented scope and stop early"), a Quick Check before deeper execution, and a fallback/error-handling path is present — more than the bare sequence of a 2. However, the workflow is generic boilerplate never connected to the script's actual modes (bridge/complete-graph/hypothesis/landscape) or flags, and duplicated sections ("Implementation Details" pointing to "See ## Workflow above") muddy the sequence, matching the 3 anchor of steps listed with implicit or generic checkpoints rather than the 4 anchor of mode-specific validation. | 3 / 5 |
Progressive Disclosure | The first References section points to `references/guide.md`, `references/examples/`, and `references/api-docs/`, none of which exist in the bundle; only the second References section's `references/audit-reference.md` is real. Combined with duplicated sections and a ~40KB main.py whose detail stays inlined via boilerplate rather than being split out, this matches the 2 anchor (references buried/dangling, content that should be separate kept inline); it is not a 3 because navigation actively misleads via nonexistent paths. | 2 / 5 |
Total | 9 / 20 Passed |