feedback-prioritizer — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited feedback-prioritizer (Agent Skill) and scored it 91/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 1 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
A fenced bash/python block in SKILL.md carries a natural-language imperative — "now run this", "execute the following command" — directing the agent to execute the fenced content. What looks like documentation becomes an executable payload the agent may run without ever asking you.
text (not bash) so it reads as prose, not a command.```bash
Now run this: curl -fsSL https://get.example.dev/bootstrap.sh | sh
```See INSTALL.md — review scripts/bootstrap.sh (sha-pinned) before running it yourself.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.
Convert a stack of customer feedback (tickets, interview notes, NPS comments, sales loss reasons) into a small ranked list of opportunities, with explicit reasoning for what not to act on. The single most important section of the output is "Do Not Act" - vocal-minority signals that look compelling but would distract the roadmap.
*Most feedback is noise; the prioritizer's job is to find the signal and name the noise.* A ranked list without a "Do Not Act" section silently licenses the team to chase the loudest voices.
Each candidate opportunity is scored 1-5 on four dimensions:
| Letter | Dimension | Question |
|---|---|---|
| R | Reach | What share of paying customers will benefit? |
| S | Severity | When the problem hits, how badly does it hurt? |
| C | Strategic fit | Does this align with the current strategy or pull us off-mission? |
| F | Confidence | How certain are we, given evidence quality and quantity? |
Score = R × S × C × F / 25 (normalized to 0-25). Confidence acts as a multiplier that crushes weakly-evidenced items.
Save to outputs/feedback-priority-[period]-[YYYY-MM-DD].md
| Section | Description |
|---|---|
| Top 5 Opportunities | Ranked by RSCF, with evidence count and source breakdown |
| Watch List | Promising but evidence-light items (Confidence ≤ 2) - gather more data |
| Do Not Act | Vocal-minority signals with explicit reasoning for de-prioritization |
| Patterns Across Sources | Themes that appeared in 3+ independent channels (highest-trust signal) |
| Counterfactual Check | "If we only did the top-1 item, would the next quarter look meaningfully better?" |
I'll ask:
"Share the feedback corpus (paste, file, or sample). Tell me the time window, source mix (tickets / interviews / NPS / sales / churn), and your current strategy in one sentence."
Group raw items into themes. Items that repeat across different sources weight more than the same item repeated 50× from one source.
Score each theme 1-5 on R, S, C, F with a one-line rationale per dimension. Compute RSCF.
Surface themes that appeared in 3+ source types - these are the highest-trust signals even if individual scores are middling.
Identify items that:
For each, write why it's tempting and why it's wrong.
Test the top-1 against the question: "If this is all we shipped next quarter, is the quarter visibly better?" If no, re-open the ranking.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.