The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell utilities to inspect the local environment and project history.
Evidence: Workflow steps include execution of git diff, git log, ls, and grep to analyze file changes and verify paths.
[INDIRECT_PROMPT_INJECTION]: The skill's design creates a surface for indirect prompt injection because it processes untrusted file content to define new agent behaviors.
Ingestion points: Changed file names and contents identified via git are used to extract patterns for new skill creation (SKILL.md files).
Boundary markers: The skill does not define specific boundaries or markers to isolate ingested content from its internal logic.
Capability inventory: The skill possesses the ability to write files to the agent's global configuration directory (~/.claude/skills/) and the project root (CLAUDE.md), and can execute arbitrary shell commands via dry-runs and detection steps.
Sanitization: The instructions do not specify sanitization procedures for data extracted from project files before it is incorporated into new agent instructions.

manage-skills