idea-engine — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited idea-engine (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
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.
Most AI tools help you write papers. This engine helps you decide which papers to write.
The current paper machine takes any topic and produces. This engine adds the missing gate: Phase 0 — Is this paper the right use of my time?
Great research starts with taste, strategic problem selection, honest self-evaluation, and knowing when to kill your darlings.
/evaluate-idea [topic]/write-paper pipeline (before Reconnaissance)Reference: principles/research-strategy.md for full details.
Goal: Gather enough context to brainstorm and evaluate intelligently.
research-evaluations/*.mdGoal: Expand the possibility space before narrowing.
agents/brainstormer.md)Goal: Honest, structured assessment of each candidate.
agents/idea-critic.md)| Dimension | Idea A | Idea B | Idea C |
|-------------|---------------|---------------|---------------|
| Novelty | Years | Months | Weeks |
| Impact | High | Medium | Low |
| Timing | Well-Timed | Well-Timed | Too Late |
| Feasibility | Medium Risk | Low Risk | Low Risk |
| Competition | Open | Moderate | Crowded |
| Nugget | Clear | Fuzzy | Clear |
| Narrative | Compelling | Workable | Weak |
| **VERDICT** | **PURSUE** | **REFINE** | **KILL** |Goal: For surviving ideas, assess strategic viability.
agents/research-strategist.md)Goal: The decisive test. Can this paper tell a compelling story?
For each surviving idea, write:
The critical test: If the conclusion feels hollow or generic — if it only says "our method achieves X% improvement" — that IS the signal. The idea doesn't have enough impact to justify months of work.
Goal: Clear, actionable decision for each idea.
For each idea, deliver one of three verdicts:
PURSUE — This is worth your time. Next steps:
PARK — Not now, but potentially later. Document:
KILL — Not worth pursuing. But still extract value:
Save evaluation to research-evaluations/YYYY-MM-DD-<topic-slug>.md:
---
date: YYYY-MM-DD
topic: [descriptive title]
verdict: [PURSUE/PARK/KILL]
nugget: [one-sentence key insight]
revisit: [date or condition, if PARK]
---
## Dimension Scores
[table from Phase 3]
## Key Concerns
[top 3 risks]
## Strategic Assessment
[summary from Phase 4]
## Draft Conclusion
[from Phase 5]
## Next Steps / Salvage
[from verdict]agents simultaneously for different ideas.
Everything else is preparation for this moment.
If the user provides a single clear idea, collapse Phases 1-2 and go straight to evaluation.
directly to Phase 1 (Reconnaissance) of the paper machine. The evaluation artifacts inform the literature search strategy.
Standalone (/evaluate-idea): Run the full 6-phase evaluation. End with verdict. The user decides whether to proceed to /write-paper.
Pipeline (Phase 0 of /write-paper): Run an abbreviated evaluation:
When Phase 0 produces a PURSUE verdict, the following artifacts carry forward:
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.