The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, potentially untrusted data to influence its output and project state.
Ingestion points: The skill uses renku screenplay import-fdx --file to ingest Final Draft files and uses renku asset list followed by various "harness readers" to ingest supporting materials such as PDFs, documents, and images.
Boundary markers: The instructions include guidelines to separate sourced facts from inventions and warn the agent not to mutate screenplays based on supporting material unless explicitly requested. However, there are no hard technical delimiters (like XML tags or multi-layer escaping) defined for the interpolated content.
Capability inventory: The skill possesses broad capabilities including executing renku CLI commands for screenplay creation/revision, writing files to the tmp/ directory, and modifying project metadata (loglines, synopses, premises) via renku info set.
Sanitization: No specific technical sanitization or filtering of the ingested content is described; the skill relies on the agent's semantic judgment and user confirmation for ambiguous identity resolution.
[COMMAND_EXECUTION]: The skill instructions heavily utilize the renku CLI for project management and screenplay operations. While these commands (renku create, renku screenplay apply, renku info set) are necessary for the skill's primary function, they represent an execution surface that could be leveraged if the agent's logic is influenced by malicious instructions embedded in the screenplays or supporting documents it is tasked to read.

screenplay-drafter