The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements a 'self-capturing learning loop' that journalizes invocations and synthesizes 'Playbooks'. These playbooks contain command sequences with variable slot substitution (e.g., Playbook.steps) that the agent is instructed to replay. This system creates a dynamic execution environment where shell commands are assembled and executed based on persistent local state, which could be exploited if the learning store is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data to generate prompts for an AI generation pipeline.
Ingestion points: Processes recap_script.json and assets imported from Google Drive.
Boundary markers: The skill does not define specific delimiters or 'ignore' instructions for the data being drafted into prompts.
Capability inventory: The skill uses flow-pp-cli to perform file writes and trigger AI generations.
Sanitization: There is no evidence of sanitization or validation for the input script content before it is transformed into ready-to-approve prompts.
[CREDENTIALS_UNSAFE]: The 'Auth Setup' section provides detailed instructions for users to manually harvest sensitive Google Bearer tokens (ya29.*) and session cookies (__Secure-next-auth.session-token) from their browser's developer tools. While these are presented as necessary configuration steps, they involve handling and exposing high-value credentials that remain valid for approximately one hour.
[COMMAND_EXECUTION]: The skill requires the Read Bash tool to execute a wide variety of commands through the flow-pp-cli binary, including operations for file muxing, project management, and generation polling.
[EXTERNAL_DOWNLOADS]: The skill references installation of the flow-pp-cli and flow-pp-mcp tools via npx and go install from the vendor's GitHub repository. These are documented as prerequisites for the skill's functionality.

pp-flow