Data & agents / Skills

Agent skills

Drop-in instruction packs that teach an agent to answer SRF accountability questions by fetching published JSON over HTTPS. Exactly one accountable party per activity. No MCP server, no API keys.

srf-accountability

Claude Code · Cowork · Agent SDK

Anthropic Agent Skills format. Download the published files into a project or user skills directory. No repository checkout required. The agent fetches live SRF JSON for accountability, vertical controls, vendor risk, threats, finding routing, Tapestry / Federated-Consortium, and glossary lookups. 

mkdir -p .claude/skills/srf-accountability
curl -fsSL -o .claude/skills/srf-accountability/SKILL.md \
  https://aisharedresponsibility.com/developers/skills/srf-accountability/SKILL.md
curl -fsSL -o .claude/skills/srf-accountability/README.md \
  https://aisharedresponsibility.com/developers/skills/srf-accountability/README.md
README SKILL.md GitHub

openclaw-srf

OpenClaw

OpenClaw-native skill folder (opt-in, not always-on). Download into the workspace skills directory, then restart the gateway. Uses built-in web_fetch against the same published files. No repository checkout required. 

mkdir -p ~/.openclaw/workspace/skills/openclaw-srf/reference
curl -fsSL -o ~/.openclaw/workspace/skills/openclaw-srf/SKILL.md \
  https://aisharedresponsibility.com/developers/skills/openclaw-srf/SKILL.md
curl -fsSL -o ~/.openclaw/workspace/skills/openclaw-srf/reference/data-sources.md \
  https://aisharedresponsibility.com/developers/skills/openclaw-srf/reference/data-sources.md
openclaw gateway restart

If you already have the folder on disk, you can also run openclaw skills install /path/to/openclaw-srf --as srf-accountability.

README SKILL.md Data sources GitHub

For paste-ready system instructions without installing a skill, use System Instructions under Tools.