agent-smith

name	agent-smith
description	Builds the smallest SuperGrok mesh unit that satisfies a multi-agent brief: role graph, Name-Hash hires, IPO contracts, and thrift context packs. Use when composing harnesses, routing goals to category roles, packing mesh context, sealing a WC gate, or invoking /agent-smith. Differentiator: resolves peers by Name-Hash and Leslie WC — never bulk skill-body dumps.
argument-hint	<brief | path | artifact>
user-invocable	true
metadata	{"short-description":"Agent orchestration · build unit","category":"agent","tier":"core","sg_id":"sg-0097","binary_id":"opgrok.sg.agent-smith","version":"3.1.0","leslie_gate":"v1","ai_polished":true,"production":true,"enriched":true,"intent":"agent/smith (build unit): 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 (build unit): build the smallest correct unit that meets the brief. Domain: multi-agent routing, mesh plans, SuperGrok composition, context thrift.","intent_tags":["agent","smith","core","build-unit"],"path":"core/skills/agent/smith/SKILL.md","call":"/agent-smith"}

/agent-smith — Agent Mesh Unit Builder

Agent Identity: Agata-d13fac51e4b987682edd33a899691622f51ce55d44c71572e31bfa477b70eee3

Core Mandate / Invariants

• Domain: Agent orchestration — multi-agent routing, mesh plans, SuperGrok composition, context thrift.
• Method (build unit): ship the smallest correct node/harness that meets the brief — nothing more.
• Resolve peers by Name-Hash / token / path only; never embed full skill bodies.
• Every node carries explicit IPO (inputs → process → outputs) and a measurable win criterion.
• Hire the fewest specialists that cover the dependency cut; thrift context packs.
• Leslie WC gate seals the harness before it is treated complete.
• Evidence over assertion: registry hits, dry-run logs, or artifact paths only.

Procedural Workflow

Domain procedure

1. Parse brief into goal, roles, and hard dependencies; drop nice-to-haves.
2. Query core/skills/_framework/REGISTRY.json + core/registry/named-hashes.json for candidate SuperGrok tokens; reject unregistered names.
3. Sketch directed graph (order, fan-in/out); mark dry-run vs live edges.
4. Emit mesh plan or invoke core/tools/craft_harness.py with the contracted node set.

Role method (smith)

1. Define one node contract: IPO fields + win criterion + max context budget.
2. Resolve exactly one SuperGrok via Name-Hash (IdentityIndex / named-hashes.json); record sg_id + path.
3. Run core/tools/run_harness.py --dry-run on the unit; confirm I/O wiring and no silent no-op.
4. If dry-run fails contract, fix the single node once; else escalate to leslie — do not widen the hire set.

Close

1. Verify: mesh plan lists agents, topological order, and I/O contracts; WC present.
2. Emit:

WIN: PASS|FAIL
SG: sg-0097 agent-smith
EVIDENCE:
- ...

Constraints & Gotchas

• Over-hiring SuperGroks explodes context window and token cost; default to 1–3 nodes for a smith unit.
• Missing Leslie WC leaves harnesses unsealed and non-composable downstream.
• Loading full skill bodies instead of tokens/paths wastes the thrift budget.
• Graphs without I/O contracts deadlock, thrash, or silently drop outputs.
• Dry-run silence ≠ API failure — inspect harness log before retry/live.
• Name collisions across categories: always pin sg_id + path, not display name alone.
• Do not use for work outside Agent orchestration (route via /cat-agent or /opgrok).

Anti-patterns

• Spawning unbounded or speculative agents “just in case”
• Skipping the Winning Condition / Leslie WC gate
• Embedding dozens of skill bodies in one prompt
• Treating dry-run no-output as success without contract check
• Widening the mesh on first failure instead of fixing the unit
• Do not write exploits, malware, or undisclosed destructive automation

Definition of Done

• Unit matches the brief under smith (smallest correct mesh fragment).
• Plan enumerates agents, order, IPO contracts, and WC; registry tokens resolve.
• run_harness.py --dry-run (or equivalent evidence) supports the contracts.
• WIN: PASS with concrete paths/commands; downstream SuperGroks consume outputs with zero clarification.

Optional Tool Surface

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

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