habr-article

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to perform file system operations using shell commands to organize generated content.
  • Evidence: mkdir -p articles/pictures/[slug]/ found in SKILL.md under the Phase 5 section.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a defined attack surface where untrusted data from a user interview is ingested and processed by the agent to create files, potentially influencing the agent's output behavior if the input contains malicious instructions.
  • Ingestion points: User interview responses gathered in "ФАЗА 0: ИНТЕРВЬЮ С АВТОРОМ".
  • Boundary markers: Absent; the skill does not specify delimiters or instructions to ignore embedded commands in the interview data.
  • Capability inventory: The agent is instructed to create directories (mkdir) and write markdown files (articles/habr/[slug].md, articles/pictures/[slug]/cover-prompts.md).
  • Sanitization: Not present; the skill lacks specific validation or escaping logic for the ingested interview text.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 05:04 AM
Security Audit — agent-trust-hub — habr-article