The Agent Skills Directory

[PROMPT_INJECTION]: The instructions establish a rule hierarchy that allows external, potentially untrusted project files to override the skill's own core instructions.
Evidence: "When a skill-context rule conflicts with a general rule written in this SKILL.md, the skill-context rule wins (more specific context takes priority)".
Evidence: "Do NOT ignore skill-context rules even if they seem to contradict this skill's defaults — they exist because the project's experience proved the default insufficient".
[COMMAND_EXECUTION]: The skill performs shell-based operations to analyze and modify the git repository.
Evidence: git status, git diff --cached, git commit, git reset HEAD, and git push.
[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill processes untrusted input from the codebase and external configuration files while possessing repository-write and network-send capabilities.
Ingestion points: Staged code changes via git diff --cached, project configuration in .ai-factory/config.yaml, and context rules in .ai-factory/skill-context/aif-commit/SKILL.md.
Boundary markers: The skill lacks explicit boundary markers or delimiters when interpolating external content into the model's context.
Capability inventory: The skill can read file contents, modify the repository state (git commit, git reset), and transmit data to remote servers (git push).
Sanitization: No sanitization or validation logic is present to filter malicious instructions within the staged diffs or the project-level context files before they are processed by the agent.

aif-commit