orca-actions — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited orca-actions (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.
Generate action scaffold code for the Orca machine.
Parse $ARGUMENTS:
typescript.If a file path is given, read the file first.
Call generate_actions with the source and lang.
scaffolds)Show all action stubs in a single fenced code block with the appropriate language tag (typescript, python, or go).
decisionTableCode)If the result contains decisionTableCode (a map of DT name → compiled evaluator code), show each evaluator in its own fenced code block, labeled with the DT name:
// Decision table: PaymentRouting
<evaluator code>Place the DT evaluator blocks after the action scaffolds.
After showing the output, briefly explain:
machine.registerAction(...) for TS, @machine.action(...) decorator for Python, machine.RegisterAction(...) for Go)When a .orca.md file contains both a machine and one or more # decision_table blocks:
apply_routing_decision ↔ PaymentRouting via "routing") — receive a commented example of how to call the evaluator, showing all input conditions and output fields with // TODO: map from ctx annotations.The user is responsible for mapping context fields to DT inputs and applying DT outputs back to context. The generated stubs provide the full evaluator signature as a guide.
Do not add boilerplate explanation beyond this. The scaffold code is self-documenting.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.