The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data in the form of the codebase being audited.
Ingestion Points: The skill reads project files, dependency lockfiles, and git history during the orienting and auditing phases (SKILL.md, references/security-checks.md).
Boundary Markers: There are no specific instructions for the agent to use delimiters or ignore embedded instructions when reading content from the audited files.
Capability Inventory: The skill has the capability to execute shell commands (git, bash, and various package managers) and write/fix files in the repository during the 'Pass 2' phase (SKILL.md).
Sanitization: The instructions do not specify any sanitization or escaping of the ingested code content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill performs shell command execution to perform its audit functions.
It uses git status, git grep, and git log to analyze the repository state and history (SKILL.md, references/security-checks.md).
It executes a bundled script (scripts/dep_audit.sh) that invokes local development tools such as npm, yarn, pip, poetry, cargo, and go to check for outdated packages and known vulnerabilities.
These executions are part of the core functionality expected for a codebase auditing tool.

anti-koshary