The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests and processes source code files from the local project codebase. As these files are untrusted external data, they represent a surface for indirect prompt injection where malicious instructions embedded in code comments or string literals could attempt to influence the agent's behavior.
Ingestion points: Production source code files identified during the 'Pre-flight' and 'Workflow' stages.
Boundary markers: Absent; the instructions do not specify delimiters or warnings to ignore instructions embedded within the files being mutated.
Capability inventory: Shell execution (running test suites via pytest, make, etc.), file modification (via the Edit tool), and Git operations (git status, git diff, git checkout).
Sanitization: Absent; the agent is instructed to read, model, and modify the source code directly.
[COMMAND_EXECUTION]: The skill executes shell commands to interact with the Git version control system and to run the project's test suite. These operations are essential to the skill's primary purpose of mutation testing.
Evidence: Instructions to run git status, git checkout -- <file>, and git diff <file> to manage code state, as well as commands to invoke test runners like pytest or tox (SKILL.md).

mutation-testing