aidlc-workflows
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/preflight.pyhelper script invokes thegitcommand to verify the local repository state. This execution is performed usingsubprocess.runwith a hardcoded list of safe arguments (rev-parse,--show-toplevel,HEAD) and a 2-second timeout to prevent hangs. The environment is restricted to a minimal set of non-sensitive variables.\n- [INDIRECT_PROMPT_INJECTION]: The skill's pre-flight script reads project-level configuration files such as.mcp.jsonand.claude/settings.json. While these files could contain malicious instructions from external sources, the skill treats the content as structured JSON or plain text for auditing purposes (e.g., counting hooks or identifying unpinned packages) rather than executing the content, which effectively mitigates the risk.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the AI-DLC tool from its official repository. While the README suggests a standard pipe-to-shell installation, the skill instructions explicitly recommend a safer download-first approach with optional attestation verification using the GitHub CLI. These downloads originate from a well-known, trusted organization repository.\n- [DATA_EXPOSURE]: The pre-flight script inspects environment variables and project local settings (like.claude/settings.local.json). However, the script is designed to only check for the presence of specific configuration keys (e.g., AWS Bedrock settings) and explicitly states that values are not read out or exfiltrated. This is consistent with its primary purpose as an offline auditing tool.
Audit Metadata