Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add NicoGenti/opencode-orchestrator-kit --skill comprehension-workflowgit clone --depth 1 https://github.com/NicoGenti/opencode-orchestrator-kitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nicogenti/opencode-orchestrator-kit/comprehension-workflow)<a href="https://agentmods.dev/skills/nicogenti/opencode-orchestrator-kit/comprehension-workflow"><img src="https://agentmods.dev/badge/skills/nicogenti/opencode-orchestrator-kit/comprehension-workflow/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nicogenti/opencode-orchestrator-kit/comprehension-workflow"><img src="https://agentmods.dev/badge/skills/nicogenti/opencode-orchestrator-kit/comprehension-workflow.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00071 | $0.03841 |
| Opus 5.5 | $0.00028 | $0.01536 |
| Sonnet 5.5 | $0.00014 | $0.00768 |
| Haiku 4.5 | $0.00007 | $0.00384 |
Grade A, and why
comprehension-workflow scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 187 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Comprehension Workflow
Detailed operational protocol for the comprehension gate. The orchestrator
loads this skill when it must run a LIGHT or DEEP evaluation, write gate
telemetry, maintain a per-evaluation record, or execute /recall. Classification
(NONE/LIGHT/DEEP), the closing rule, and escalation routing stay in
agents/orchestrator.md.
Configuration and precedence
The comprehension gate reads its limits from a user-local config file: .opencode/comprehension.config.json (user copies it manually from templates/comprehension.config.json; install.sh does not install templates/).
Precedence: file value > built-in defaults. If the file is absent, empty, or not valid JSON → all limits fall back to the built-in defaults listed below, and the user receives ONE warning line in the progress notes (never a hard failure).
Invalid values (e.g. negative questions, missing required keys) are detected by the orchestrator's validator and treated as invalid → the offending entry falls back to its default, with a one-line warning logged.
Default limits:
| Key | Default |
|---|---|
enabled |
true |
mode |
adaptive |
light.questions |
2 |
light.modelTier |
TIER_FAST |
deep.questions |
4 |
deep.modelTier |
TIER_FAST |
deep.escalationTier |
TIER_REVIEW |
context.maxDiffLines |
300 |
context.unifiedContextLines |
3 |
context.maxRelatedSymbols |
3 |
evaluation.maxRetries |
1 |
evaluation.escalateOnAmbiguity |
true |
The orchestrator reads these limits when composing delegations to comprehension-coach.
enabled: false skips the gate entirely — orchestrator proceeds straight to close and records comprehension=DISABLED in the progress notes.
Tier resolution stays on the existing TIER_FAST / TIER_REVIEW mapping via the preset resolver.
Context slicing (Slicer)
Trigger — the Slicer activates ONLY when the estimated change size exceeds context.maxDiffLines (default 300) or when the number of changed files requires a bounded slice. Tasks classified NONE receive 0 coach calls (no slice needed). Small changes follow the unchanged v0.3.0 path (specialist change report + coach reads the listed files only).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 187 lines · 71 tokens per session scan A 476e0f6e7900
comprehension-workflow is a skill published in the GitHub repository NicoGenti/opencode-orchestrator-kit (7 stars, last pushed 5d ago), licensed MIT. It adds 71 tokens to every session and 3,841 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-08.
Other skills, from other repositories
cao-workflow
Author and run CAO Python workflow scripts — multi-step, parameterized, fan-out orchestrations executed by cao workflow run. Use when the user wants a repeatable multi-step job (e.g. data analysis over many files, a review pipeline, a parameterized batch). Authoring ends at a validated script file; running it is a…
aidlc-portfolio
Coordinate multiple AI-DLC workflows across repositories and Git worktrees using an evidence-backed portfolio catalog and deterministic workspace tooling. Use when initializing an AI-DLC portfolio workspace, discovering organization or business context, registering projects and dependencies, creating child intents and…
cao-plugin
Create a new CAO (CLI Agent Orchestrator) plugin. Use this skill whenever the user wants to add a plugin that reacts to CAO lifecycle or messaging events, scaffold a plugin package, understand plugin requirements, or integrate an external system (Discord, Slack, dashboards, logging, metrics) with CAO. Also use when…
cao-mcp-apps
Enable, operate, and extend CAO's MCP Apps surface — the host-rendered fleet dashboard visible inside MCP App hosts (Claude Desktop, ChatGPT, VS Code Copilot, Goose, Postman). Use when the user says "enable MCP Apps in CAO", "the ui://cao views aren't rendering", "rebuild MCP Apps bundles", "add a new ui://cao/ view"…
cao-learning
Report task outcomes and distill lessons so the team improves across runs — reportoutcome after each unit of work, retrospector handoffs at natural boundaries, and applying injected lessons. Use in workflows that run repeatedly over similar work items. Requires memory.learningenabled; degrade silently when the tools…
cao-supervisor-protocols
Supervisor-side orchestration patterns for assign, handoff, and idle inbox delivery in CAO.