investigate

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill is strictly read-only and explicitly prohibits making changes to source code, plans, or specifications. It enforces a rigorous evidence-based reasoning protocol, requiring the agent to cite file:line for every finding and mark unverified claims as inferred.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because its primary purpose is to ingest and summarize untrusted data from a codebase (source code, logs, and documentation).
  • Ingestion points: The skill reads arbitrary project files, git diffs, and project-specific reference documentation.
  • Boundary markers: It utilizes a structured 'Knowledge Graph' template and mandatory todo tracking to maintain task boundaries, although it lacks explicit instructions to ignore embedded instructions within the code it reads.
  • Capability inventory: The skill is capable of reading files, searching with grep, and executing local repository scripts (Python and Node.js).
  • Sanitization: The instructions do not specify sanitization or escaping of code content before it is processed or summarized.
  • [DYNAMIC_EXECUTION]: The skill invokes existing local utility scripts located in the .claude/scripts/ directory, such as code_graph (Python) and ai-signal-scan.cjs (Node.js). These tools are used for repository analysis tasks like tracing dependencies and mapping API connections.
  • [COMMAND_EXECUTION]: The skill frequently uses shell commands (e.g., grep, python, node, git) to explore the codebase. These commands are directed at the local repository environment and are necessary for its documented purpose of code investigation.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 08:16 PM