press

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a bundled Python archive at skills/press/scripts/press.pyz for routing, telemetry, and report generation. These operations are internal to the skill's workflow and occur within its own scripts directory.\n- [DYNAMIC_EXECUTION]: As part of its adversarial testing mission, the skill executes test suites and fixtures (e.g., via pytest). This dynamic execution is the primary intended function of the skill for verifying contract compliance.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests handoff artifacts and glossary terms from previous agent phases, creating a potential surface for indirect instruction injection.\n
  • Ingestion points: Reads .cheese/cook/<slug>.md and .cheese/glossary/<slug>.md during phase entry.\n
  • Boundary markers: None explicitly defined for the ingested Markdown content.\n
  • Capability inventory: Shell command execution via press.pyz and the ability to run test suites.\n
  • Sanitization: The skill instructions do not specify sanitization or validation logic for the content of the handoff artifacts.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 04:47 AM
Security Audit — agent-trust-hub — press