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 companion-inc/introspect --skill agent-control-channelsgit clone --depth 1 https://github.com/companion-inc/introspectWrote 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/companion-inc/introspect/agent-control-channels)<a href="https://agentmods.dev/skills/companion-inc/introspect/agent-control-channels"><img src="https://agentmods.dev/badge/skills/companion-inc/introspect/agent-control-channels/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/companion-inc/introspect/agent-control-channels"><img src="https://agentmods.dev/badge/skills/companion-inc/introspect/agent-control-channels.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 | $0.00104 | $0.01494 |
| Opus 5 | $0.00052 | $0.00747 |
| Sonnet 5 | $0.00021 | $0.00299 |
| Haiku 4.5 | $0.00010 | $0.00149 |
Grade A, and why
agent-control-channels 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 10d 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 — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent control channels
The one rule
When the model must signal a structured decision to the host, that signal goes through a native structured channel — a tool call for actions/control, or a strict structured response schema for pure data that the app will render. It is not a marker or prompt-only JSON object the model types into its reply.
A typed marker (an XML-ish tag, a sentinel string, <<QUIET_RUN>>, a magic word) fails three ways at once:
- It leaks. Anything the model types into its reply is part of the reply. It streams to the user verbatim. "Why is it leaking?" is the symptom; the typed marker is the cause.
- It's emitted unreliably. The model stutters it, half-types it, rewords it, or wraps it in prose — because it's just more text being sampled, with no schema holding its shape.
- It forces brittle recovery. The host now regex-scrapes the prose to recover intent (
extractXFromParts). That parser is permanently chasing the model's phrasing.
A tool call has none of these: it's a separate structured field, schema-validated, never rendered into the user-facing transcript, and reliably shaped because the API constrains it. The model invoking archive() cannot accidentally show the user the word "archive."
Prompt-only JSON is the same smell
Return only JSON is not a control channel. It is prose with braces. It is acceptable only as a fallback when the provider or runtime truly lacks tool calls and structured outputs.
For a menu, routing choice, suggested action list, app mode, or other host-consumed payload:
- use a tool/function call when the model is choosing an app action, changing state, invoking code, or handing the host a command;
- use a strict structured response schema when the model is only returning data that will be rendered, such as a UI card, menu, or explanation parts;
- keep the host parser as schema validation, not recovery from a natural-language prompt contract.
Do not ship a prompt that says "return only JSON" while the host JSONDecoders or JSON.parses the assistant text when the runtime supports one of those native channels.
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.
- 10d ago First seen · 74 lines · 104 tokens per session scan A cd6393476f25
agent-control-channels is a skill published in the GitHub repository companion-inc/introspect (10 stars, last pushed 22d ago), licensed MIT. It adds 104 tokens to every session and 1,494 once invoked, about $0.0005 per session on Opus 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 2026-08-31.
Other skills, from other repositories
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
local-ai-agents
Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…
next-cache-components-adoption
Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…
insight-error-page
Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…
next-cache-components-optimizer
Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…
next-partial-prefetching-adoption
Turn on Partial Prefetching in a Next.js app and work through the insights it surfaces. Use when the user wants to enable or adopt Partial Prefetching, flip the partialPrefetching flag, opt routes in with export const prefetch = 'partial', audit Link prefetch={true} behavior, preserve existing prefetched UI with…