mojo

Use when working on the Sub-agent Dispatch harness and you need capability control: Registers new agent types, pluggable planning strategies, and enforced result schemas.

Use when working on the Orchestrator harness and you need capability control: Lets operators swap the "why this next" decision logic (round-robin, risk-first, deadline-first) without touching the orchestrator core.

Use when working on the Trace harness and you need performance control: Writes traces asynchronously so audit logging doesn't block the call path.

Use when working on the Orchestrator harness and you need cost efficiency control: Caps concurrent high-cost reasoning sessions per project and reclaims idle slices, holding total attention spend under a set ceiling.

Use when working on the Latency harness and you need capability control: Adds SLO config and availability probes for new routes.

Use when working on the Provider harness and you need cost efficiency control: Chooses batch over realtime endpoints and compresses payloads where latency allows to reduce vendor cost.

Use when working on the Latency harness and you need cost efficiency control: Trades latency for cheaper batch processing where the task tolerates delay.

Use when working on the Cost harness and you need performance control: Checks budget before dispatch so over-budget work is stopped before it spends.

Use when working on the Prompt harness and you need cost efficiency control: Maximizes prompt-cache hits and strips redundant context to cut input tokens.

Use when working on the Network Protocol harness and you need cost efficiency control: Prefers local execution, budgets remote calls, and uses degraded-mode fallbacks to limit cross-node cost.

Use when working on the Tools harness and you need cost efficiency control: Caps tool calls per request so a run can't overspend on tool use.

Use when working on the Tools harness and you need performance control: Caches results and runs safe tool calls in parallel to reduce wait.

Use when working on the Fallback harness and you need capability control: Defines the fallback chain and retry policy for new failure modes.

Use when working on the Browser Capture harness and you need cost efficiency control: Captures only on detected change and reuses browser sessions to reduce screenshot and compute cost.

Use when working on the Identity harness and you need cost efficiency control: Pins to the cheapest sufficient model version when capability allows, lowering per-call cost.

Use when working on the Knowledge Retrieval harness and you need cost efficiency control: Tries cached embeddings and cheap recall before broad search, bounding retrieval breadth to the context budget.

Use when working on the Git Control harness and you need cost efficiency control: Minimizes branch-lease scope and gates CI triggers (rebase-vs-merge policy) to avoid redundant pipeline cost.

Use when working on the Git Control harness and you need performance control: Precomputes likely merge conflicts and tunes shallow-fetch and diff-risk scoring so Git operations plan faster.

Use when working on the Memory harness and you need capability control: Adds a memory schema, conflict-resolution logic, and provenance depth so recall stays trustworthy under contradiction.

Use when working on the Policy harness and you need capability control: Lets teams author policy and install constraint packs for new route/task rules.

Use when working on the Context harness and you need performance control: Compacts context and selects the right window size so prompts fit with less processing.

Use when working on the Goal harness and you need cost efficiency control: Drops context not tied to the active objective to save tokens.

Use when working on the Cost harness and you need capability control: Adds per-provider cost models and budget-policy authoring.

Use when working on the Proof Authority harness and you need performance control: Runs evidence crosschecks across lanes and caches prior proofs so sufficiency scoring returns faster.

Use when working on the Response harness and you need cost efficiency control: Stores deltas and compresses artifacts to shrink response storage cost.

Use when working on the Guardrails harness and you need performance control: Limits secret and permission scans to changed surfaces and precompiles rule sets so gate evaluation stays fast.

Use when working on the Project Scope harness and you need performance control: Recomputes only changed scope regions and caches the dependency graph so drift checks stay cheap and current.

Use when working on the Tool Registry harness and you need cost efficiency control: Enforces tool-call ceilings, dry-runs costly actions first, and reuses sandboxes to avoid wasted spend.

Use when working on the Compare harness and you need performance control: Runs comparison lanes in parallel and exits on early consensus to save time.

Use when working on the Evidence harness and you need performance control: Gathers evidence across sources in parallel to shorten linking time.

Use when working on the Evidence harness and you need capability control: Adds connectors and a link schema so output ties to new validation sources.

Use when working on the Evidence harness and you need cost efficiency control: Scales evidence depth to risk tier so low-risk output isn't over-evidenced.

Use when working on the Fallback harness and you need performance control: Sets tight fail timeouts and keeps a pre-warmed standby so blocked routes recover quickly.

Use when working on the Identity harness and you need performance control: Pins requests to the fastest adequate model variant to trim response time.

Use when working on the Review Queue harness and you need performance control: Collapses duplicate findings and batches clusters across parallel reviewer lanes so the queue clears faster.

Use when working on the Goal harness and you need performance control: Caches the goal binding so each call reattaches to the objective without recompute.

Use when working on the File Navigation harness and you need performance control: Builds the ownership graph incrementally, caps impact-walk depth, and caches paths so navigation stays fast on large trees.

Use when working on the Safety harness and you need performance control: Caches guard evaluations so repeated safe patterns clear quickly.

Use when working on the Guardrails harness and you need capability control: Lets teams author secret patterns, approval-gate definitions, and credential-boundary policy as installable packs.

Use when working on the Network Protocol harness and you need performance control: Caches trust handshakes and batches signature verification so multi-node calls add less per-packet latency.