The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to structure and report on external research data, which presents an inherent surface for indirect prompt injection.\n
Ingestion points: External research data and URLs provided during the research phase, processed into the markdown report in SKILL.md.\n
Boundary markers: The skill mandates the use of ## TL;DR sections and explicit citation markers [n] which act as structural delimiters.\n
Capability inventory: The skill utilizes a local Python script measure.py for file analysis. The agent follows markdown structuring rules.\n
Sanitization: The skill relies on formatting conventions and a validation script to ensure output consistency and clear separation of source material.\n- [COMMAND_EXECUTION]: The skill includes a helper script references/measure.py and instructions in references/checklists.md for its execution. A technical audit of the script confirms it uses only standard Python libraries (argparse, re, statistics, sys, pathlib) for regex-based text analysis. It contains no network calls, shell execution functions, or obfuscated logic.\n- [SAFE]: The core logic of the skill consists of instructional markdown files that guide the agent through report generation phases and styles without requesting unnecessary permissions or using risky tools.

survey-report