The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains no instructions designed to override agent behavior, bypass safety filters, or extract system prompts. The "Session Protocol" section establishes legitimate operational boundaries for managing the screenwriting process.
[DATA_EXFILTRATION]: No network operations (curl, wget, fetch) or exfiltration patterns were identified. The skill reads and writes to local project files (story-bible.md, .fountain, .txt), which is entirely consistent with its purpose as a writing project manager.
[OBFUSCATION]: Analysis found no Base64 encoding, zero-width characters, homoglyphs, or hidden text patterns. Terminology tables in terms.md are correctly mapped and use legitimate scripts.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute external code. It references other internal skills (e.g., sw-premise-theme, sw-story-structure) for specific craft tasks, which is standard modular design for AI agent skills.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from a local story-bible.md file, which tracks user-provided project details. While this is an entry point for external data, the skill uses this information solely for literary context and state management. It lacks dangerous capabilities like network access or arbitrary command execution that would allow for an exploitable injection attack.

sw-workflow