The Agent Skills Directory

[SAFE]: The skill implements a robust framework for scientific evidence synthesis, referencing reputable academic sources such as the Cochrane Handbook, PRISMA 2020 guidelines, and the JBI Manual. All external links point to official documentation domains (e.g., prisma-statement.org, cochrane.org, osf.io).
[COMMAND_EXECUTION]: The package includes a local validation script scripts/self_check.py. This script performs static analysis on the skill's own files to ensure integrity and check for the presence of hardcoded private data (emails, user paths, or API keys). It does not perform network operations or execute code derived from user input.
[SAFE]: The skill explicitly defines 'Source Discipline' and 'Privacy Minimization' policies in SKILL.md and references/atoms.jsonl. These policies prohibit the storage of sensitive user information (names, emails, tokens, or absolute paths) within the skill's code or generated outputs.
[SAFE]: Indirect Prompt Injection surfaces are managed by instructions that require the agent to clearly separate user-provided materials from public sources and its own inferences. The skill's primary function is report and template generation, lacking capabilities for high-risk operations like arbitrary shell execution or unauthorized file system modifications.

rw-review-methods