The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill requires the agent to generate Python source files and test modules, then execute them using tools like pytest and python -c to verify implementation. This represents the execution of locally generated content.
Evidence: SKILL.md workflow steps 4, 6, 7, and 8; TESTING.md workflow sections.
[INDIRECT_PROMPT_INJECTION]: The workflow involves ingesting user tasks and questions to generate code and tests that are subsequently executed in the agent's environment. There is a potential attack surface where a malicious prompt could result in the generation and execution of unintended code during the verification steps.
Ingestion points: SKILL.md frontmatter lists task and question as input types.
Boundary markers: Not specified in the generation workflow.
Capability inventory: Shell execution via pytest, pip, and python as described in SKILL.md and TESTING.md.
Sanitization: No explicit sanitization or validation of the generated code before execution is mentioned.
[COMMAND_EXECUTION]: The workflow relies on executing shell commands to manage the environment, install the package, and run tests.
Evidence: Use of pip install -e, pytest -v, and python -c for installation and public API verification.
[METADATA_POISONING]: The skill metadata lists static-analysis under capabilities, which is inconsistent with the workflow's reliance on dynamic execution (running tests and installing packages).
Evidence: SKILL.md frontmatter capabilities vs implementation instructions.

build-python-library