The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external markdown files (brain-storming documents) and existing specification files to generate new project specs.
Ingestion points: The skill uses Glob and Read tools to access brain-storm/**/*.md and specs/**/*.md based on user-provided titles or paths.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when interpolating content from these external files into the final spec templates.
Capability inventory: The skill has access to powerful tools including Bash, Write, and SendMessage.
Sanitization: There is no mention of sanitizing, escaping, or validating the content read from external files before it is used to seed new specification sections, creating a risk that malicious instructions in those files could be executed by the agent.
[DYNAMIC_EXECUTION]: The skill instructs the agent to run a Python script using a dynamically computed path, which is a technique that can be used to execute code outside the immediate scope of the skill.
Evidence: The skill includes a bash command template: python3 "$(readlink -f ~/.claude/skills/writing-specs)/../check-doc-structure/check_doc_structure.py" <repo> --roadmap.
Reasoning: The use of readlink -f to resolve paths at runtime for script execution falls under dynamic execution from computed paths, which increases the difficulty of auditing the specific code being run.

writing-specs