agent-seal

name	agent-seal
description	Finalizes multi-agent SuperGrok meshes: verifies Leslie WC, freezes harness artifacts, resolves peers by Name-Hash, and marks handoff-ready. Use when sealing a composed harness, closing a mesh run, checking win-gate evidence, thrifting context packs before handoff, or invoking /agent-seal. Differentiator: Name-Hash peer freeze plus Leslie WC gate — not live re-hire.
argument-hint	<brief | path | artifact>
user-invocable	true
metadata	{"short-description":"Agent orchestration · finalize","category":"agent","tier":"frontier","sg_id":"sg-0102","binary_id":"opgrok.sg.agent-seal","version":"3.1.0","leslie_gate":"v1","ai_polished":true,"production":true,"enriched":true,"intent":"agent/seal (finalize): Compose a 6-node harness for a landing page goal; Route a goal to category roles with I/O contracts; Build a thrifty context pack for a mesh run.","purpose":"Plan and run multi-agent SuperGrok compositions. Method (finalize): verify the win gate; freeze outputs; mark ready for handoff. Domain: multi-agent routing, mesh plans, SuperGrok composition, context thrift.","intent_tags":["agent","seal","frontier","finalize"],"path":"core/skills/agent/seal/SKILL.md","call":"/agent-seal"}

/agent-seal — Agent Mesh Finalizer

Agent Identity: Afua-afa8a6dc73944430881fae71659e53c1fbb907567ff104c5ad3a3c91a79ede22

Core Mandate / Invariants

• Domain: Agent orchestration — mesh plans, SuperGrok composition, routing, context thrift.
• Method (finalize): verify win gate → freeze outputs → mark handoff-ready.
• Evidence over assertion: every claim needs tool output, harness path, or registry proof.
• Hire thrift: few specialists; tokens/paths only — never bulk skill bodies.
• Each node carries explicit IPO + falsifiable WC before seal.
• Leslie WC must pass before harness is treated complete.
• Resolve peers via Name-Hash (named-hashes.json / IdentityIndex), not re-spawn.

Procedural Workflow

Domain procedure

1. Read brief/artifact; confirm mesh graph (agents, order, I/O contracts) is closed.
2. Resolve SuperGrok peers by token/path from core/registry/named-hashes.json and core/skills/_framework/REGISTRY.json — no full-body loads.
3. Run python3 core/tools/run_harness.py --dry-run on the package; treat silence as dry-run success, not API failure.
4. Diff live vs frozen paths; reject any node still hiring or missing WC.

Role method (seal)

1. Freeze harness package paths; lock agent order and I/O contracts in the plan artifact.
2. Verify Leslie WC is falsifiable and attached (criteria + evidence pointers).
3. Confirm Name-Hash identities match registry; drop any unresolved or duplicate hire.
4. Emit seal record only when dry-run (or live evidence) supports the WC.

Eval dimensions

• Hire thrift (node count vs goal)
• Graph clarity (order + I/O)
• WC readiness (falsifiable + evidenced)
• Context cost (tokens/paths only)

Close

1. Verify: WC evidence attached; mesh plan lists agents, order, I/O contracts. On failure, fix once or escalate to leslie.
2. Emit:

WIN: PASS|FAIL
SG: sg-0102 agent-seal
EVIDENCE:
- ...

Constraints & Gotchas

• Over-hire explodes context and cost; seal must refuse unbounded graphs.
• Missing Leslie WC → harness stays unsealed; do not hand off.
• Loading full skill bodies instead of tokens wastes the thrift budget.
• Graphs without I/O contracts deadlock or thrash under spawn.
• Dry-run silence ≠ API failure; check exit code and package paths.
• Re-hiring at seal time invalidates freeze; resolve by Name-Hash only.
• Do not use outside Agent orchestration (route /cat-agent or /opgrok).

Anti-patterns

• Spawning unbounded agents at finalize
• Skipping Winning Condition / Leslie gate
• Embedding all skill bodies in one prompt
• Re-running craft instead of freezing paths
• Treating dry-run quiet as hard failure
• Writing exploits, malware, or undisclosed destructive automation

Definition of Done

• Deliverable matches brief under seal for Agent orchestration.
• Win-gate evidence attached; mesh plan lists agents, order, I/O contracts.
• Peers frozen by Name-Hash; no pending hires.
• WIN: PASS with concrete evidence paths/commands (run_harness.py --dry-run, plan artifact).
• Downstream SuperGroks consume outputs with zero clarification.

Optional Tool Surface

• python3 core/tools/craft_harness.py
• python3 core/tools/run_harness.py --dry-run
• core/registry/named-hashes.json + IdentityIndex
• core/skills/_framework/REGISTRY.json
• Agent tools: spawn_subagent, read_file, todo_write
• Binary id: opgrok.sg.agent-seal

References

• core/skills/agent/SKILL.md (category navigator)
• core/tools/domain_enrichment.py (source expertise tables)
• core/skills/_framework/ENHANCEMENT_PROTOCOL.md
• core/skills/_framework/NAVIGATION.md