ctx-doctor
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill defines a fallback routine that calculates a file path relative to its own installation directory and executes the resulting file (
cli.bundle.mjsorcli.js) using thenoderuntime. Dynamic loading and execution from computed paths can be used to run arbitrary code. - [COMMAND_EXECUTION]: The fallback logic uses a Bash shell string to perform file existence checks and execute the Node.js CLI bundle.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests output from the
ctx_doctorMCP tool and a local CLI, instructing the agent to display these results verbatim to the user. This creates a vulnerability surface where malicious content in the tool output could influence the agent's behavior. - Ingestion points: MCP tool output and CLI stdout defined in
SKILL.md. - Boundary markers: None provided; the skill explicitly requests verbatim rendering.
- Capability inventory: Invocations of MCP tools and execution of shell commands (Bash/Node).
- Sanitization: None detected.
Audit Metadata