The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a restricted Bash tool to execute a local project script named yaml-helper.sh. The tool definition in the skill's frontmatter limits the command to this specific script path to prevent arbitrary command execution.
[DYNAMIC_CONTEXT_INJECTION]: The skill employs the ! command syntax to run a shell command at the time the skill is loaded. Specifically, it executes bash "${CLAUDE_SKILL_DIR}/../../hooks/yaml-helper.sh" resolve_config to retrieve project settings such as rigor levels and certification targets. This is used for legitimate project-specific configuration resolution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by scanning source code files for comments (TODO, FIXME, HACK) and including their content in the final checklist.
Ingestion points: Source files are read using the Glob and Grep tools.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious instructions embedded within the scanned code comments.
Capability inventory: The skill has the ability to write files to the production/releases/ directory and execute a specific local bash script.
Sanitization: The skill does not perform sanitization or escaping of the scanned comment text before it is presented in the markdown report.

release-checklist