The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The PromptTemplate.render method in SKILL.md implements a simple string substitution mechanism (result.replace(f"{{{key}}}", str(value))) that facilitates the inclusion of external data into agent prompts.
Ingestion points: The render method accepts arbitrary kwargs that are placed directly into template placeholders.
Boundary markers: The provided templates (e.g., agent_base.yaml, code_review.yaml) use header markers like ## Task but do not include explicit delimiters or instructions for the LLM to ignore embedded commands in the interpolated data.
Capability inventory: The skill itself focuses on string manipulation and local file reading, but the prompts it generates are used to control the agent's behavior across all its available tools.
Sanitization: There is no evidence of validation, escaping, or filtering of the content provided in the kwargs before interpolation.
[SAFE]: The implementation uses yaml.safe_load() for parsing template files, following best practices to prevent unsafe deserialization vulnerabilities.
[SAFE]: No network operations, hardcoded credentials, privilege escalation attempts, or persistence mechanisms were detected in the provided Python implementation or YAML examples.

prompt-template