The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands to discover the project environment and manage state.
Evidence: Uses git rev-parse --show-toplevel to identify the project root and test -f to check for existing configuration files (SKILL.md).
Evidence: Invokes ccw spec init and ccw spec rebuild to manage local project specifications.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted local project data to inform its setup process.
Ingestion points: Analyzes project structure and files (e.g., package.json, requirements.txt, go.mod) through a subagent (cli_explore_agent) to determine the technology stack and architecture (SKILL.md).
Boundary markers: The instructions provided to the subagent lack explicit delimiters or instructions to ignore potential commands embedded within the project files being analyzed.
Capability inventory: The skill has access to powerful tools including Bash, Write, Edit, and spawn_agent, which could be abused if the analysis phase is compromised.
Sanitization: There is no evidence of sanitization or validation of project-derived content before it is passed to the analysis agent.

spec-setup