The Agent Skills Directory

[SAFE]: The skill exclusively fetches data from authoritative Japanese government and research agency websites (cao.go.jp, mext.go.jp, meti.go.jp, jsps.go.jp, jst.go.jp, amed.go.jp, nedo.go.jp, etc.), which are considered highly trusted entities.- [SAFE]: Tool usage is limited to standard operations; the Bash tool is used for benign tasks like date calculation and data formatting, and the Write tool is used for saving markdown reports.- [SAFE]: The skill does not contain any obfuscated code, credential exposure, or persistence mechanisms. Its behavior aligns perfectly with its stated purpose of policy monitoring.- [SAFE]: Regarding the potential indirect prompt injection surface: data ingestion occurs via WebFetch from official domains in SKILL.md. While explicit boundary markers are absent, the high trust level of the source domains and the lack of high-risk command execution ensure the process is safe. Capability inventory is restricted to local file writing and basic formatting.

policy-monitor