skills/rootspec/demos/rs-spec/Gen Agent Trust Hub

rs-spec

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes several local bash scripts (e.g., scan-spec.sh, validate-spec.sh, write-stats.sh) to perform project analysis and validation. These scripts are located in a relative path within the vendor's shared directory.- [PROMPT_INJECTION]: The skill implements an 'Indirect Prompt Injection' surface by reading all project source files to automatically derive specifications. In its 'non-interactive mode', the agent skips user review and drafts content directly from code context, which could allow malicious instructions embedded in code comments to influence the final documentation. Evidence of ingestion: Step 1 (Assess) reads source code; Step 5b (Reconcile) reads files to match acceptance criteria. Capability inventory: Script execution for scanning and writing file status. Sanitization: Not explicitly mentioned. Boundary markers: No explicit markers are defined for the source code ingestion.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 08:24 AM
Security Audit — agent-trust-hub — rs-spec