agent-runtime-validati…
Plans and evaluates bounded runtime tests of AI agents, tools, memory, hooks, approvals, and scope enforcement. Use when static…
Validates supplied immutable ordered manifests of opaque defensive assessment actions for authority, safeguards, evidence, and separate-operator review. Use when reviewing assessment sequencing without creating, changing, or executing actions.
Claude Code
$ npx -y skills add zero-lab-ai/skills --skill assessment-sequence-governance --agent claude-code
Codex
$ npx skills add zero-lab-ai/skills --skill assessment-sequence-governance -a codex -y
Cursor
$ npx skills add zero-lab-ai/skills --skill assessment-sequence-governance -a cursor -y
Gemini CLI
$ npx skills add zero-lab-ai/skills --skill assessment-sequence-governance -a gemini-cli -y
GitHub Copilot
$ npx skills add zero-lab-ai/skills --skill assessment-sequence-governance -a github-copilot -y
OpenCode
$ npx skills add zero-lab-ai/skills --skill assessment-sequence-governance -a opencode -y
Windsurf
$ npx skills add zero-lab-ai/skills --skill assessment-sequence-governance -a windsurf -y
name: assessment-sequence-governance description: "Validates supplied immutable ordered manifests of opaque defensive assessment actions for authority, safeguards, evidence, and separate-operator review. Use when reviewing assessment sequencing without creating, changing, or executing actions." allowed-tools: Read Grep Glob
Validate a supplied ledger without designing the assessment. A passing result means only that the recorded sequence is internally consistent for separate review. It is never execution authority.
Never create, select, add, remove, reorder, rank, or substitute actions or dependencies. Never choose targets, routes, tactics, timing, or tools. Never schedule or run anything. Do not provide commands, queries, configurations, credentials, payloads, protocols, infrastructure, stealth, or deployment guidance. Accept opaque identifiers and digests only.
1. Obtain the already-supplied immutable ledger and confirm the operator wants an offline governance review. 2. Stop if the material contains operational instructions or live identifiers. Do not quote the rejected value. 3. Preserve the supplied manifest and action order byte-for-byte. Do not repair missing actions, dependencies, bindings, or evidence. 4. Reject malformed shapes, types, references, timestamps, digests, restricted content, duplicate identifiers, and an empty assessment universe. 5. Keep manifest integrity separate from action identity, authority, prerequisites, environment, budget, stop, cleanup, and custody coverage. 6. Require exact action, sequence, manifest, scope, environment, policy, authority, and window joins. Similar-looking records never substitute. 7. Preserve supplied predecessor sets. Verify acyclicity, contiguous ordinals, and predecessor chronology; do not improve or reorder the sequence. 8. Treat budgets as nontransferable reservations with nonnegative integer attempts, duration, impact, and evidence-byte dimensions. 9. Require a passed stop pretest, an armed available stop, and exact action-set binding. Bind canonical budget, stop, and cleanup fingerprints through the authority and every supporting evidence record. 10. Accept cleanup `not-required` only with exact current policy proof. Never infer cleanup from silence. 11. Count every action against all nine required coverage roles. Never omit a blocked, prohibited, or historical action. 12. Derive the gate in precedence order: `prohibited`, `approval-required`, `blocked`, then `sequence-consistent`. 13. Treat a recorded handoff as historical custody of review work, not permission. Derive `withdrawal-required` after authority expiry or revocation, manifest or action drift, budget exhaustion, a fired or lost stop, cleanup loss, or custody loss. 14. Accept `withdrawn` only from a later exact withdrawal receipt. Never infer retry, resumption, or restored authority.
State the manifest identifier, derived gate, handoff state, contradictions, coverage gaps, and withdrawal obligations. Add: “Internal consistency only; separate review is required; no execution authority is granted.”
manifest.
the supplied sequence.
consistent for separate review only.
Portable, evidence-bound security skills for authorized defensive assessment, penetration testing, red teaming, detection engineering, purple teaming, and engagement operations. The collection contains 56 skills in 10 installable plugins from zero-lab.ai.
Plans and evaluates bounded runtime tests of AI agents, tools, memory, hooks, approvals, and scope enforcement. Use when static…
Reviews an MCP server's tool surface for tool poisoning, over-broad parameters, missing descriptions, and privilege escalation…
Maps identity, authority, instruction, data, memory, and approval propagation across agents and agent-to-agent handoffs. Use when…
Assesses whether direct, indirect, multi-turn, retrieved, or tool-mediated instructions can override an AI system's intended…
Maps cloud identity and authorization paths across AWS, Azure, GCP, and hybrid trust boundaries, separating proven access from…
Tracks embedded-device and firmware assessment coverage across exact units, acquisition methods, debug interfaces, stored…
Use for authorized security assessment of REST, GraphQL, WebSocket, or SOAP APIs, including discovery, authentication,…
Extract and decrypt Windows DPAPI-protected secrets (Credential Manager, browser logins/cookies, Wi-Fi credentials, KeePass keys)…
Clean and finalize authorized natural-language text intended for readers by auditing suspicious invisible Unicode and rewriting…