The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs several shell operations to analyze the codebase, including running git diff to identify modified files and using rg (ripgrep) or grep to trace function callers and imports.
[COMMAND_EXECUTION]: It dynamically locates a local Python script named graph_query.py within the ~/.claude/plugins directory and executes it to perform graph-based impact analysis.
[COMMAND_EXECUTION]: The skill executes a citation verifier script (plugins/imbue/scripts/citation_verifier.py) to validate that the findings correspond to verbatim source code anchors.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it processes untrusted data from git diff and file contents throughout the project (SKILL.md Step 2) without explicit boundary markers or sanitization. This is a common characteristic of code analysis tools.
Ingestion points: Reads output from git diff and file contents via rg and grep in Step 2.
Boundary markers: None are explicitly used for the ingested code content.
Capability inventory: The skill can execute shell commands, read files, and run local Python scripts.
Sanitization: No evidence of input sanitization or filtering was found for the data read from the codebase.

blast-radius