apc-test

Warn

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill directs the agent to load instructions from a relative path (../../.agents/workflows/apc-test.md) that resides outside the skill's dedicated directory. This pattern constitutes dynamic loading of instructions from a computed path, meaning the skill's actual logic and behavior are deferred to an external file not included in the analyzed skill package.
  • [INDIRECT_PROMPT_INJECTION]: The mechanism of loading external instructions creates a significant surface for indirect prompt injection, as the agent is instructed to adopt the external file's content as authoritative.
  • Ingestion points: The file SKILL.md contains a command to ingest and follow the content of ../../.agents/workflows/apc-test.md.
  • Boundary markers: Absent; the agent is explicitly instructed to follow the external file "exactly as written," providing no delimiters or warnings to ignore potentially malicious embedded instructions.
  • Capability inventory: The skill description implies the agent will perform build verification, DAW (Digital Audio Workstation) checks, and interface with the pluginval utility.
  • Sanitization: Absent; there is no evidence of validation, integrity checking, or sanitization of the external file's content before it is processed by the agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 19, 2026, 01:17 AM