The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is inherently designed to ingest and analyze untrusted data from an external repository, creating a significant attack surface for indirect prompt injection.
Ingestion points: The skill reads and processes the target repository's source code, CLAUDE.md, documentation, MEMORY.md, and git log history (documented in SKILL.md and analisi.md).
Boundary markers: The orchestration prompts in repo-audit.js do not include explicit delimiters or instructions for sub-agents to ignore potentially malicious content within the analyzed files.
Capability inventory: The skill has access to shell execution (for tests/builds/audits), file system read access, and the ability to spawn additional sub-agents via the agent() function.
Sanitization: There is no evidence of content sanitization or filtering before repository data is passed to the AI agents.
[COMMAND_EXECUTION]: The repo-audit.js script identifies and executes shell commands defined within the analyzed project's metadata (e.g., package.json, Makefile). While the script contains instructions to avoid destructive commands and use temporary environments, this remains a high-capability risk if the agent is influenced by malicious project-defined scripts.
[DYNAMIC_EXECUTION]: The skill employs a JavaScript orchestration file (repo-audit.js) to manage the audit workflow. It dynamically constructs prompts and tasks sub-agents to perform various analysis phases, which is an expected but high-complexity execution pattern.

dido-code-style