proposal-review — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited proposal-review (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Produce a decision-ready review of an AI/ML, computational biology, or bioscience proposal. Be fair, skeptical, and specific, and name missing information.
Strong Accept (>=4.5), Accept (>=3.7), Borderline (>=2.8), or Reject (<2.8). A documented fatal flaw may override the numeric band. uv run --script "$HOME/.agents/skills/proposal-review/scripts/score_proposal.py" scorecard.jsonWith the default rubric, scores must contain exactly these 1-to-5 keys: strategic_fit_and_novelty, technical_rigor, feasibility_and_resources, team_and_execution, risk_ethics_and_compliance, budget_and_schedule. For a sponsor rubric, add "rubric": {"weights": {...}, "recommendations": [{"minimum": 4, "label": "Fund"}, ...]}; weights must total 100 and the score keys must match them. Set "fatal_flaw": true (and optionally fatal_flaw_recommendation) to override the band. The helper rejects weights that do not total 100, category mismatches, and scores outside 1 to 5, and never mixes default bands into a sponsor rubric.
| Task | Action |
|---|---|
| Summarize proposal | Describe aims, novelty, and bottom-line recommendation in <=150 words |
| Test strategic logic | Answer the Heilmeier catechism explicitly |
| Review feasibility | Check assumptions, methods, milestones, and resource realism |
| Review rigor | Assess controls, baselines, validation, statistics, and reproducibility |
| Review risk | Build a risk register with likelihood, impact, warning signs, and mitigations |
| Make a decision | Give a final recommendation plus concrete funding conditions or rejection reasons |
| Use a sponsor rubric | Preserve its categories, weights, thresholds, and recommendation labels |
| Validate a scorecard | uv run --script "$HOME/.agents/skills/proposal-review/scripts/score_proposal.py" scorecard.json |
Review this proposal for a microbiome foundation-model project. Use a 1-5 scorecard,
identify fatal flaws if any, and list conditions for funding.Review this translational bioscience proposal for a program with a 24-month timeline,
$1.5M budget cap, and high concern for regulatory risk.Issue: The proposal is missing a clear evaluation plan Solution: Mark this as a major weakness, explain what convincing evidence would look like, and add PI questions about milestones and success metrics.
Issue: The budget or timeline is hard to judge Solution: State the uncertainty, identify the likely critical path, and evaluate whether the claimed scope is credible under the stated constraints.
Issue: Ethics or compliance details are absent Solution: Treat the omission as a potential blocker and ask targeted questions about subjects, privacy, biosafety, or regulatory readiness.
/manuscript-review-council: equivalent review for manuscripts/scientific-writing: draft or revise the proposal narrative/bio-logic: assess methodology and evidence rigor~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.