The Agent Skills Directory

[PROMPT_INJECTION]: The skill metadata uses authoritative directives ('REQUIRED
Always activate FIRST', 'DO NOT IGNORE') to override the agent's default tool selection logic and prioritize its instructions.\n- [PROMPT_INJECTION]: Indirect Prompt Injection Surface:\n
Ingestion points: The skill instructs the agent to read repository history and commit content using jj log, jj show, and jj diff (SKILL.md).\n
Boundary markers: No delimiters or safety instructions are defined to separate the untrusted repository data from the agent's internal instructions.\n
Capability inventory: The agent has access to jj via Bash(jj *), which enables both local file mutations and network operations like git fetch and git push (SKILL.md).\n
Sanitization: The skill lacks sanitization or validation logic for the content retrieved from the VCS history.

jujutsu