The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from repository files during the commit planning phase, creating a surface for indirect prompt injection.
Ingestion points: git diff, git status, and git diff --cached are used to read file changes as described in Step 1 and 3 of SKILL.md.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are used when the agent processes file contents or diff outputs.
Capability inventory: The skill can execute shell commands via the git CLI, including state-changing operations like add, commit, and switch.
Sanitization: There is no evidence of sanitization or filtering of the file content before it is processed by the agent to generate messages and plans.
Mitigation: The skill implements a mandatory human-in-the-loop approval gate for all commit plans, requiring the user to authorize any proposed actions before they are executed.
[COMMAND_EXECUTION]: The skill uses the git CLI for repository management. These operations are executed locally and are consistent with the skill's stated purpose as a developer tool.

krt-gitflow-knight