The Agent Skills Directory

[PROMPT_INJECTION]: No direct attempts to override system instructions, bypass safety filters, or extract system prompts were detected in the skill instructions or metadata.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill does not access sensitive system paths (such as .ssh, .aws, or .env) or perform network operations to exfiltrate data. It operates on local input provided by the user and generates markdown files.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: There are no external package installations or remote script downloads identified in the files.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data via the --from-example [file] parameter defined in SKILL.md.
Ingestion points: Untrusted content enters the context when the agent reads a user-provided rubric file to create a variant.
Boundary markers: No explicit delimiters or instructions to ignore embedded instructions within source files are present.
Capability inventory: The skill is limited to markdown generation and does not have capabilities for command execution, network access, or sensitive file system operations.
Sanitization: No input validation or escaping of the ingested file content is performed before analysis.
[OBFUSCATION]: No encoded content, multi-layer obfuscation, or deceptive character use was found.
[SAFE]: The rubric templates and meta-framework provided in templates/templates-library.md follow professional standards for assessment validity and reliability without exhibiting malicious behavior.

rubric-creator