Skip to content

opensiro/vsm-harness-skills / skills/assess-vsm-harness

assess-vsm-harness

Verified

Produce a standalone evidence-backed VSM assessment of one autonomous AI agent harness at a pinned repository revision, including first-party ownership modes.

references/interventions.md

View on GitHub
# Functional intervention methods

Use this guide after a boundary-aware assessment has identified a consequential organizational gap. The methods describe functions and relationships before implementation mechanisms.

## Independent S3* audit

1. Name the operational claim whose truth matters.
2. Trace how the claim is produced, transformed, reported, and used by S3.
3. Identify correlated dependencies: evidence sources, prompts, models, credentials, metrics, code paths, and owners.
4. Select access to operational reality that bypasses enough of that path for the stated risk.
5. Define sampling or trigger conditions, freshness, authority to inspect, disagreement handling, and coverage limits.
6. Route findings to S3 without allowing ordinary control to capture the audit path.

State the independence argument explicitly: claim, ordinary path, audit path, shared dependencies, remaining correlated-failure risk, recipient, and authority. Logs, tests, or a second evaluator count only to the extent that they can reveal discrepancies the ordinary path could hide.

## Exception and algedonic escalation

1. Establish the S1 unit's local purpose, authority, capability, and viable bounds.
2. Identify exceptions exceeding local authority, capability, time horizon, or risk allowance.
3. Route each exception to the nearest recursion level with requisite information and legitimate authority.
4. Define evidence, urgency, maximum tolerable delay, interim containment, acknowledgement, resolution, and feedback.
5. Reserve algedonic bypass for exceptional pain or opportunity that cannot tolerate normal reporting compression.
6. Test suppression, storms, duplicates, unavailable authorities, stale decisions, and gamed signals.

No severity vocabulary, message shape, or transport is required by the profile.

## Recursive organization

A candidate recursive unit needs a durable parent-level contribution, distinct environment and outcomes, meaningful local variety, local identity, autonomy, and the metasystemic functions needed for its viability.

1. Draw current recursion levels and assign each material decision to one level.
2. Identify variety currently centralized that the candidate could absorb better.
3. Define local autonomy and parent constraints in plain language.
4. Design the candidate's internal viable relationships proportionate to its complexity.
5. Define accountability, negotiation, and escalation across the parent boundary.
6. Test duplicate S3 control, conflicting S5 identities, audit capture, and coordination overhead.

If the candidate lacks these properties, represent it as delegation or technical composition rather than VSM recursion.

## S3–S4 balance

1. Map S3's current capability model: commitments, resources, constraints, performance, and operational risks.
2. Map S4's external/future model: signals, scenarios, threats, opportunities, experiments, and adaptation options.
3. Trace both directions of their conversation and the latency or distortion in each.
4. Identify dominance, absent variety, incompatible measures, or missing decision rights.
5. Locate the S5 identity and policy criteria that close genuinely unresolved tension.
6. Propose a measurable rebalancing experiment rather than renaming an internal planner as S4.

## Variety management

1. State the regulator, regulated domain, viable bounds, and available response time.
2. Enumerate disturbance classes and distinctions that change the correct action.
3. Trace those distinctions through sensing, filtering, summarization, routing, decision, action, and feedback.
4. Mark attenuation, amplification, transduction, delay, and shared bottlenecks.
5. Test whether attenuation removes only noise or hides control-relevant distinctions.
6. Test whether amplification adds usable response capacity or merely adds agents, context, messages, and cost.
7. Move locally absorbable variety toward autonomous S1 units and escalate only what exceeds their bounds.

## Intervention record

For any method, report:

```text
gap · evidence · variety regulated · authority changed · proposed mechanism
success signal · correlated failure · rollback condition · unresolved uncertainty
```

Prefer the smallest reversible change compatible with the existing harness. Do not infer a new protocol, runtime, or mandatory agent topology.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall