ablation-planner

Warn

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill generates experiment scripts and configuration files at runtime based on its analysis of method descriptions and component lists found in the project. These generated scripts are then executed to perform the ablation studies.
  • [COMMAND_EXECUTION]: The workflow involves using the Bash tool to run the generated scripts and smoke tests during the implementation phase (Step 5).
  • [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes data from external project files and includes it in prompts for a secondary agent.
  • Ingestion points: Reads content from research_contract.md, EXPERIMENT_LOG.md, EXPERIMENT_TRACKER.md, and other project notes in Step 1.
  • Boundary markers: None; the gathered context is interpolated directly into the prompt for the gpt-6-astra model without delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill has access to Bash, Write, Edit, Read, Grep, and Glob tools.
  • Sanitization: No sanitization or validation of the content read from project files is performed before it is used to generate the ablation plan.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 10, 2026, 09:36 AM