The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses various shell commands to manage the state of the repository and handle backups of the current work.
Evidence includes calls to git branch, git diff, git ls-files, git restore, tar, mktemp, and xargs rm -f in Step 4.
It also instructs the agent to execute arbitrary project-specific commands like tests, linters, and type checks in Step 3 and Step 6.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it processes untrusted data and possesses significant shell capabilities.
Ingestion points: Reads the project's source code and the output from external checks (tests, linters).
Boundary markers: None; the skill does not wrap the ingested code or check outputs in protective delimiters or instructions to ignore embedded commands.
Capability inventory: Can execute Git commands, create/modify directories, delete files via rm, and run arbitrary shell commands defined as project checks.
Sanitization: None; the skill does not validate or sanitize the output of the checks or the content of the code before processing it.

redoing-implementations