The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run within scripts/git_evidence.py to execute git log and git rev-parse commands. The revision range arguments are strictly validated against injection patterns (e.g., checking for leading dashes and ensuring proper range formatting) to prevent command injection.
[DYNAMIC_EXECUTION]: The skill executes Python scripts located in the project's subdirectories ({project-root}/_bmad/scripts/). These scripts are part of the BMad framework's customization and configuration system and are used to resolve project-specific settings. This is a standard functional requirement for vendor-owned resources in this toolchain.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from git commit messages, story files, and architecture documents. To mitigate the risk of the agent being misled by malicious content in these files, the instructions mandate that every finding must be sourced with a specific reference (file, line, or commit) and that unverified claims must be discarded.
[SAFE]: The skill uses ruamel.yaml for round-trip YAML processing in scripts/sprint_status.py, which is a secure method for handling configuration files without executing arbitrary code. The use of yaml.load() flagged by static analysis is a false positive as it utilizes the safe RoundTripLoader. Additionally, file writes are performed atomically using a temporary file and os.replace to prevent data corruption.

bmad-retrospective