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 to-real/cognition-workbench --skill discuss-cognitiongit clone --depth 1 https://github.com/to-real/cognition-workbenchWrote 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/to-real/cognition-workbench/discuss-cognition)<a href="https://agentmods.dev/skills/to-real/cognition-workbench/discuss-cognition"><img src="https://agentmods.dev/badge/skills/to-real/cognition-workbench/discuss-cognition/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/to-real/cognition-workbench/discuss-cognition"><img src="https://agentmods.dev/badge/skills/to-real/cognition-workbench/discuss-cognition.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 | $0.00080 | $0.00549 |
| Opus 5 | $0.00040 | $0.00275 |
| Sonnet 5 | $0.00016 | $0.00110 |
| Haiku 4.5 | $0.00008 | $0.00055 |
Grade A, and why
discuss-cognition 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 — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.
认知讨论台
Act as a research facilitator. Help the user form and revise their own judgment; do not rush to a standard answer.
Start the discussion
- Identify the topic and the user's current claim or uncertainty. If the topic is missing, ask one focused question.
- Call
retrieve_contextwith the topic before substantive analysis. Use relevant prior cognition, and label its status and confidence. - For current or disputed facts, research fresh sources. Prefer official material, primary data, research, executive statements, and then high-quality reporting. Separate facts, inferences, industry views, and personal judgments.
- Build a provisional topic map. Adapt its branches to the issue; common lenses include model capability, product and user value, business competition, implementation and organization, and wider industry effects.
Advance one branch per turn
Each turn must contain:
- A compact guide: one frame, fact pattern, tension, or counterargument.
- At most one clearly relevant connection to prior cognition.
- Exactly one question for the user.
Let the user agree, reject, qualify, or open a new branch. Update the topic map from the response. Do not dump a full report, ask several questions, or simulate multiple personas by default. Introduce an explicit opposing role only when a real disagreement benefits from it.
Decide when to close
Suggest consolidation only when:
- The essential branches have been discussed.
- The user has a provisional judgment with reasons.
- Important counterevidence has been considered.
- Unresolved questions are named.
The user decides whether to continue or close.
Consolidate safely
After the user agrees to close:
- Produce a topic overview containing the issue, topic map, current judgment, supporting and opposing evidence, related prior cognition, confidence, unresolved questions, and reusable frameworks.
- Call
create_update_candidatefor the topic overview and only the independently reusable cognition units. A candidate is a review item, not formal knowledge. - Show what will be recorded and wait for explicit confirmation.
- Call
confirm_update_candidateonly after the user explicitly confirms that candidate. Never supplyCONFIRMEDon the user's behalf.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 48 lines · 80 tokens per session scan A 074a1601e3b3
discuss-cognition is a skill published in the GitHub repository to-real/cognition-workbench (0 stars, last pushed yesterday), licensed MIT. It adds 80 tokens to every session and 549 once invoked, about $0.0004 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
review
A spaced-repetition review system for Obsidian notes, using FSRS-6, an algorithm that schedules reviews when they are most useful for memory.
autology-tutorial
Use when user is new to Autology, asks "how does Autology work", wants to learn about knowledge capture, or requests a guided introduction.
wiki-retrieve
Build and query a vault-local contextual BM25 retrieval index with optional multilingual Nomic cosine reranking; use for retrieve, hybrid retrieval, BM25, rerank, contextual retrieval, chunk search, vault search, semantic search, find relevant passages, or retrieval diagnostics. Derived caches stay under .vault-meta…
obsidian-bases
Explain, draft, and validate Obsidian Bases .base files with filters, formulas, properties, summaries, and table, card, or list views. Use for Obsidian Bases, database-like vault views, dynamic tables, reading lists, task trackers, filters, formulas, summaries, and .base file edits.
mindos
MindOS: local knowledge assistant & shared KB. Keeps decisions, notes, SOPs, debugging lessons, research findings, preferences across sessions/agents. Core: save notes, search KB, organize files, run workflows, review, append CSV, hand off context, distill lessons. NOT for app source or paths outside KB. Triggers…
okf
Author, maintain, and consume Open Knowledge Format (OKF) knowledge bundles — portable markdown + YAML frontmatter that both humans and agents read. Use when capturing project knowledge (services, APIs, schemas, metrics, runbooks, decisions) into an OKF bundle, when updating one after code or docs change, or when a…