The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data which creates an injection surface. Ingestion points: plan.md, spec.md, and project-level CLAUDE.md/AGENTS.md files (SKILL.md, Resolve the plan). Boundary markers: Absent. Capability inventory: git commit, git worktree, and arbitrary shell command execution via 'checks', 'gates', and 'probes' (SKILL.md, Task loop, Detours and stops). Sanitization: Absent.\n- [COMMAND_EXECUTION]: The skill frequently executes shell commands, including git status, git commit, git worktree, and repository-specific test suites (SKILL.md, Review, Task loop). These commands are used to manage the development lifecycle.\n- [DYNAMIC_EXECUTION]: The skill defines and uses 'probes', which are short, throwaway scripts generated and executed at runtime to verify code behavior (VOCABULARY.md, Oracle; SKILL.md, Detours and stops). These scripts are executed from the OS temporary directory.\n- [PERSISTENCE]: The 'Close the loop' phase appends information to project configuration files such as AGENTS.md, CLAUDE.md, or CONTEXT.md (SKILL.md, Close the loop; ARTIFACT-HOME.md, Effort-scoped). These files influence agent behavior in subsequent sessions.

lite-execute