The Agent Skills Directory

[SAFE]: The skill is composed of documentation and instructional Markdown files (SKILL.md and several reference files). It serves as a methodology for an AI agent to evaluate the quality of tests using five specific checks: Name-to-Assertion Alignment, Meaningful Assertions, Mutation Failure Check, Edge Case Coverage, and Mock Hygiene.
[REMOTE_CODE_EXECUTION]: No patterns for remote code execution, external downloads, or unpinned dependencies were found. The metadata.json confirms no external requirements or toolchains.
[DATA_EXFILTRATION]: There are no commands or instructions that attempt to access sensitive files (like .ssh or .env) or perform network exfiltration. The skill operates within the context of the files provided to it by the agent.
[PROMPT_INJECTION]: The skill contains no instructions aimed at bypassing AI safety guidelines or overriding system prompts. It uses standard instructional language to define a process for code review.
[DYNAMIC_CONTEXT_INJECTION]: No use of dynamic context patterns (such as shell execution during skill load) was detected in the SKILL.md or other files.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to ingest and analyze untrusted data (user-provided test and implementation files), it does not possess high-risk capabilities like arbitrary command execution or network access that could be exploited via indirect injection. It acts as an analytical layer for reporting purposes.
Ingestion points: SKILL.md Step 1 and 2 (Reading test files and implementation code).
Boundary markers: Absent. The instructions do not explicitly tell the agent to wrap untrusted content in delimiters or ignore embedded instructions.
Capability inventory: File reading and report generation (natural language output). No subprocess, execution, or network capabilities are requested or used.
Sanitization: Absent. The skill relies on the underlying LLM's standard processing of input data.

test-quality-inspector