skill-creator-flagos

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill's 'Improve' and 'Validate' modes require the agent to read and process the contents of other SKILL.md files and their supporting scripts. This creates an attack surface where a malicious skill being analyzed could attempt to override the agent's behavior.
  • Ingestion points: The agent reads SKILL.md, scripts/, and references/ of any skill targeted for improvement or validation.
  • Boundary markers: The instructions do not specify any delimiters or 'ignore' markers to separate untrusted file content from the agent's primary instructions.
  • Capability inventory: The skill possesses powerful capabilities including Bash execution, file modification (Write/Edit), and permission management (chmod).
  • Sanitization: There is no evidence of sanitization or filtering of the content ingested from the target skill files.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute local Python scripts (such as init_skill.py, validate_skills.py, and run_eval.py) and to set file permissions using chmod. These commands are generally well-scoped to the repository's internal tools.
  • [DYNAMIC_EXECUTION]: The skill generates new SKILL.md templates and directory structures through the init_skill.py script. While the templates are statically defined within the script, the workflow involves creating and eventually executing new sets of instructions and code files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 08:43 AM
Security Audit — agent-trust-hub — skill-creator-flagos