The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands to interact with the local environment, specifically for Git operations (git status, git diff) and to run validation tools defined in the project configuration (e.g., package.json scripts, Makefiles). This is a core function of the skill but involves executing local scripts that could be modified by an attacker in a compromised repository.
[INDIRECT_PROMPT_INJECTION]: The agent is instructed to ingest and analyze potentially untrusted data from the repository, such as source code, diffs, and configuration files. This creates a vulnerability where malicious instructions embedded in code comments or data files could influence the agent's behavior.
Ingestion points: Reading of git diff output and source code files during the "Discover context" and "Find cleanup candidates" phases in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or clear separation between the skill's instructions and the untrusted data being processed.
Capability inventory: The agent has the ability to write to the file system (to perform cleanup) and execute shell commands (for validation).
Sanitization: The skill does not provide guidelines for sanitizing or escaping the content of the files it reads before processing them.
[EXTERNAL_DOWNLOADS]: The documentation suggests installing the skill using npx skills add looties-io/looties-skills. This command fetches and executes content from the vendor's repository. As this targets the official vendor resource for looties-io, it is consistent with standard installation practices for this platform.

code-cleanup