# Agentsec Probe Suite

> Add the AgentSec probe library (200+ adversarial probes — injection, PII exfil, refusal evasion, destructive-tool-intent) to CI for pre-prod red-teaming. PLANNED — not yet implemented.

- Skill: `theagentos/agentsec-probe-suite` (Agent Skill)
- Install (CLI): `npx skillmds@latest add theagentos/agentsec-probe-suite`
- Raw SKILL.md: https://api.skillmd.com/api/skills/theagentos/agentsec-probe-suite/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: theagentos (https://skillmd.com/u/theagentos)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/theagentos/agentsec-probe-suite

---


# AgentSec probe suite — PLANNED

This skill is planned but not yet written. The intended scope:

1. Pick a probe pack (`owasp-llm-top10`, `pii-exfil`, `destructive-tool-intent`,
   `refusal-evasion`, or `all`).
2. Wire `agentsec probe` as a CI job (GitHub Actions / GitLab CI / circle.yml).
3. Configure the agent under test — endpoint, auth, sample inputs.
4. Set a pass threshold (default: ≥95% probe pass rate on the reference
   suite, matching the AgentOS v1 success metric).
5. Surface probe failures as CI annotations + AgentHog `security.alert` events.

Track progress at https://github.com/TheAgentOS/agentos-skills/issues