ai-act-skills: Instructions file for Gemini CLI

GEMINI.md — ai-act-skills

Activation pointer for Gemini CLI. Gemini reads this file at session start and uses it to decide which agent skills are available in the current working directory.

Skill exposed to Gemini CLI

Skill ID	Entry	Activation tool	Manifest
ai-act-compliance	skills/ai-act-compliance/SKILL.md	activate_skill	skills/ai-act-compliance/ssl.json (SSL-1.0, arXiv:2604.24026)

Description — EU AI Act (Regulation 2024/1689) compliance skill — risk classification, conformity assessment, Annex IV documentation, FRIA, AI literacy (art. 4), substantial modification (art. 25), regulatory sandboxes (arts. 57–63), GPAI (arts. 51–55), serious- incident reporting (art. 73), ISO/IEC 42001:2023 (AIMS) and ISO/IEC 27090:2025 (AI cybersecurity) mapping.

Activation contract

Gemini CLI should:

1. Load this GEMINI.md at session start (default behaviour).
2. Surface ai-act-compliance as available. The skill metadata is in the SKILL.md YAML frontmatter; the canonical trigger source is the description field.
3. When the user's message matches one of the intent signatures (full list in skills/ai-act-compliance/ssl.json intent_signature), call activate_skill with skill_id: "ai-act-compliance".
4. Follow the skill's typed scene graph (S_PREPARE_SCOPE → S_ACQUIRE_FACTS → S_REASON_TIER → S_ACT_OBLIGATIONS → S_VERIFY_ARTIFACTS → S_FINALIZE_REPORT, with S_RECOVER_INCIDENT as a parallel branch when the user reports a serious incident under art. 3(49)).

Tool-name compatibility

Skills built for Claude Code typically use Claude-specific tool names (Bash, Read, Edit, Glob, Grep, WebFetch). Gemini CLI ships its own equivalent set. This skill needs no translation because it declares (in ssl.json):

"control_flow_features": {
  "branching": true,
  "loops": false,
  "tool_calls": false,
  "touches_sensitive_resources": false
}