The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the Bash tool to create directories for output files using mkdir -p (Step 10). This operation uses paths derived from user-provided parameters (output_path). This is a standard functional requirement for file management within the workspace and is performed by the agent as part of the documented workflow.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because its primary function is to process untrusted narrative documents provided by users (identified in Step 0). These inputs are used to generate headlines, copy, and image prompts.
Ingestion points: Narrative markdown files searched via Glob in Step 0.
Boundary markers: Not explicitly defined during prompt interpolation, but the workflow includes structural markers in the final web-brief.md output.
Capability inventory: Write (web-brief.md), Bash (mkdir), and AskUserQuestion.
Sanitization: The skill implements a robust validation framework in Step 9 and a stakeholder review process in Step 9b to verify content integrity and schema compliance, effectively mitigating the risk of processing malicious inputs.
[SAFE]: The skill follows established development patterns for the cogni-work ecosystem, referencing internal libraries and tools like the theme picker and review assessor. No obfuscation, unauthorized network access, hardcoded credentials, or persistence mechanisms were detected.

story-to-web