run-spike
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from user requests and existing project files to generate logic models and TUI shells.
- Ingestion points: SKILL.md instructions specify deriving the spike's purpose from 'the request' and 'surrounding code'.
- Boundary markers: Absent. There are no instructions to use delimiters or 'ignore' blocks for ingested text.
- Capability inventory: The agent is authorized to write files and execute shell commands via the project's task runner.
- Sanitization: Absent. No logic is provided to sanitize or verify the safety of code generated from untrusted context.
- [COMMAND_EXECUTION]: The skill requires the agent to execute shell commands to run the generated logic spikes.
- Evidence: SKILL.md Rule 2 and LOGIC.md Step 5 explicitly instruct the agent to register and run the spike code using the project's task runner.
- [DYNAMIC_EXECUTION]: The skill functions by generating new scripts and immediately executing them to provide feedback to the user.
- Evidence: LOGIC.md describes the process of building a TUI and logic module at runtime for immediate interactive use.
Audit Metadata