The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains no instructions to bypass safety guidelines, override system prompts, or engage in unrestricted roleplay.
[DATA_EXFILTRATION]: No sensitive file access or network exfiltration patterns were detected. Network interaction is restricted to local loopback addresses (127.0.0.1) for the JSON control channel and MCP server bridge.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or private secrets were found in the skill or metadata.
[OBFUSCATION]: The content is provided in clear text with no evidence of Base64 encoding, zero-width characters, homoglyphs, or other techniques to hide malicious intent.
[INDIRECT_PROMPT_INJECTION]: The skill correctly identifies that media files and project metadata are untrusted inputs. It provides the agent with a verification protocol to check for offline media, exit codes, and output sync to prevent accidental processing of malicious or malformed data.
[COMMAND_EXECUTION]: The skill describes legitimate use of a project-specific CLI (filmcraft-cli) for headless video processing tasks, which is the stated purpose of the skill.

filmcraft