The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes untrusted data in the form of Git diffs from the current branch. Malicious content within the diff (such as comments or code patterns) could potentially influence the agent's behavior during the narrative planning or commit description phases.
Ingestion points: Branch diff analysis as specified in the execution steps (SKILL.md).
Boundary markers: None present to distinguish untrusted diff content from the agent's core instructions.
Capability inventory: Execute shell commands via git (branch, commit, push, reset).
Sanitization: No explicit sanitization or validation of the diff content is mentioned.
[COMMAND_EXECUTION]: The skill instructs the agent to execute various Git commands. It specifically mandates the use of --no-verify for intermediate commits, which bypasses client-side Git hooks (such as secret scanners or linters), and uses --force-with-lease to modify remote branch history. It also instructs the agent to avoid adding attribution metadata such as 'Generated with Claude Code'.

rewrite-history