The Agent Skills Directory

[SAFE]: The skill is a legitimate orchestrator for content preparation. It identifies clear inputs, stages, and outputs consistent with its stated purpose. Analysis of the instructions reveals no attempts at prompt injection, obfuscation, or persistence.\n- [DATA_EXPOSURE]: The skill requests access to local source materials (source_path) and git repositories (repo_path). This data access is essential for the primary functionality of summarizing content and performing 'git archaeology' for technical talks. No exfiltration patterns or unauthorized data access were found.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from local files and git history, which represents a potential attack surface for indirect prompt injection. However, this processing is the intended purpose of the tool, and the workflow includes a checkpoint requiring human confirmation before final script generation.\n
Ingestion points: source_path, repo_path (SKILL.md)\n
Boundary markers: Absent\n
Capability inventory: Read and Write tools used to process materials and save outputs (SKILL.md)\n
Sanitization: Not explicitly specified in the orchestration logic.\n- [COMMAND_EXECUTION]: The skill uses user-provided variables (slug, date) to construct file paths for writing outputs. This pattern is standard for the skill's directory management and does not include execution of arbitrary shell commands.

talk-pipeline