software-factory

Pass

Audited by Gen Agent Trust Hub on Oct 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes bash scripts (init-wo-execution.sh, update-context-index.sh) designed to automate project setup and context management.
  • These scripts utilize standard utilities such as sed, awk, and printf to perform placeholder substitution and file manipulation within a local .sw-factory/ directory.
  • The scripts implement basic safety practices, including input escaping for sed operations and the use of set -euo pipefail for robust error handling.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process external project documentation, such as Requirements Documents and Technical Blueprints, which are ingested into the agent's context.
  • Ingestion points: External documentation is read via MCP tools as specified in execute-work-order.md and review-phase.md.
  • Boundary markers: The instructions do not define specific technical delimiters (e.g., XML tags) for the content of these documents, relying instead on high-level checklist certification steps.
  • Capability inventory: The skill has the capability to write to the local filesystem and execute internal bash scripts.
  • Sanitization: Shell scripts perform variable escaping to prevent command injection during templating.
  • [EXTERNAL_DOWNLOADS]: The review-phase.md guide recommends the use of a well-known browser automation tool from a reputable source.
  • It references the ChromeDevTools/chrome-devtools-mcp repository on GitHub for browser-based testing.
  • The skill suggests that agents may set this tool up automatically when browser behavior is relevant to the implementation task.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 2, 2026, 07:06 PM