Skip to content

seb1n/awesome-ai-agent-skills / agent-security/agent-red-teaming

agent-red-teaming

Verified

Plan, execute, document, and retest authorized security assessments of AI agents and multi-agent workflows using safe adversarial cases, synthetic identities, canaries, and evidence-based findings. Use when defining red-team rules of engagement, assessing prompt injection or excessive agency, testing tool and identity boundaries, evaluating memory or cross-agent attacks, scoring a campaign, or verifying remediation in an approved environment.

FileKindSize
SKILL.mdskill md10.9 KB
agents/openai.yamlconfig246 B
assets/campaign-plan-template.jsonconfig1.7 KB
assets/campaign-plan-template.mdmarkdown2.6 KB
references/test-taxonomy.mdmarkdown5.3 KB
scripts/score_campaign.pyexecutablescript21.7 KB

assets/campaign-plan-template.md

View on GitHub
# Authorized agent red-team campaign

## Authorization and rules of engagement

- Target owner:
- Authorization status, reference, approver subject, approval time, and expiry:
- Exact target IDs, environment, version, and configuration SHA-256 digest:
- Test window and stable tester subject IDs:
- Allowed techniques:
- Prohibited systems, data, techniques, and effects:
- Rate, time, and cost ceilings:
- Stop conditions:
- Emergency contact and channel:
- Evidence location, access, retention, and deletion:
- Cleanup owner and deadline:

If authorization is pending, label this plan non-executable.

Mirror the approved fields and case IDs in [campaign-plan-template.json](campaign-plan-template.json). The JSON plan is the authority used by `score_campaign.py`; prose notes cannot add executable scope.

## System and privilege map

- Models and orchestration:
- Inputs and retrieval:
- Memory:
- Tools and external systems:
- Identities, scopes, tenants, and approvals:
- Protected assets and security invariants:
- Output sinks:

## Test matrix

| Approved case ID | Hypothesis/category | Target / environment / config digest | Preconditions | Protected invariant | Safe case/oracle | Expected safe behavior | Per-case limits | Cleanup required |
|---|---|---|---|---|---|---|---|---|
|  |  |  |  |  |  |  |  |  |

## Execution log

| Case ID / test ID | Authorization reference | Tester | Target / environment / config digest | Started / ended / recorded | Observed limits | Outcome / invariant held | Structured evidence IDs | Cleanup status / evidence |
|---|---|---|---|---|---|---|---|---|
|  |  |  |  |  |  |  |  |  |

## Findings

For each finding record title, approved case ID, unique test ID, affected target/configuration, protected invariant, severity rationale, prerequisites, minimum safe reproduction, expected/observed behavior, structured evidence, root control failure, containment, durable remediation, owner, and regression test. A failed invariant cannot be informational; non-failed outcomes do not carry finding severity.

## Retest and closure

- Findings retested:
- Nearby variants:
- Benign utility and false positives:
- Open or accepted risk and owners:
- Synthetic records removed:
- Test credentials revoked:
- Jobs/callbacks stopped:
- Fixtures restored:
- Unexpected effects or incident reference:

Report attempted and observed percentages against every approved plan case, including missing and explicitly not-run cases in the denominator. Do not label a submitted-record ratio as campaign coverage, and do not describe the result as proof of security or certification.
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall