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 wiber/thetacog-mcp --skill steer-everygit clone --depth 1 https://github.com/wiber/thetacog-mcpWrote 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/wiber/thetacog-mcp/steer-every)<a href="https://agentmods.dev/skills/wiber/thetacog-mcp/steer-every"><img src="https://agentmods.dev/badge/skills/wiber/thetacog-mcp/steer-every/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/wiber/thetacog-mcp/steer-every"><img src="https://agentmods.dev/badge/skills/wiber/thetacog-mcp/steer-every.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 | $0.00140 | $0.01078 |
| Opus 5.5 | $0.00056 | $0.00431 |
| Sonnet 5.5 | $0.00028 | $0.00216 |
| Haiku 4.5 | $0.00014 | $0.00108 |
Grade A, and why
steer-every 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 8d 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 — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/steer-every — one steer file, round after round, a settle between
This is the STRANGER-SAFE /steer-every. It ships inside the ThetaCog npm package and the VS Code extension, and every command below is a door: npx thetacog-mcp <door>. A door not listed here does not exist here yet; say not run — no <name> door rather than guessing.
The doors
| what it does | door |
|---|---|
| run the loop in the background | npx thetacog-mcp steer-every <steer-file> --detach [--every-min 5] [--max-rest-min 20] [--runs 12] [--budget-usd 7.50] [--ceiling 0.75] |
| is a loop running, and where is it | the last lines of <steer-file>.runs.log (a rest line last means between rounds) and pgrep -fl steer-every |
| one settle, now (reap, unload local models, rest until calm) | npx thetacog-mcp mem --settle [--min 5] [--max 20] |
| what is holding memory, kills nothing | npx thetacog-mcp mem |
| write a steer file from a dictation | npx thetacog-mcp steer-compile "<dictation>" |
The shape (learned 2026-10-28, keep it)
- The file comes first. A steer file carries one line
SCOPE: C1,C2,…(the rows each round works). Rows are added to the file; a track keeps one file. SCOPE is re-read every round, so promoting a row needs no restart. - No chat in the loop. Each round is one
steer until --max-cycles 1with turn-capped headless workers. A chat-scheduled tick re-reads the whole conversation every time; this loop spends only what its workers spend. - A round, then a settle. After every round the loop calls
mem --settle: reap what nobody is coming back for, unload every model ollama holds, then rest at least--every-min, on until CALM, never past--max-rest-min. CALM is both: 1-min load per core at most 0.4, and under 100 swap-ins/s. So the period is round time plus rest, by design, never a fixed slot. - Swap used is never waited on. It sits at 85–97 % on a busy Mac, calm or not, and shrinks only when the processes holding its pages exit. The settle line names the biggest holders; stop the ones that are yours and idle (a dev server nobody is using), reload the apps.
- The budget is projected before a round. The loop stops when spent plus its largest round so far would pass
--budget-usd. Every round so far has run past its per-worker ceiling, so the ceiling alone never caps the spend.
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.
- 8d ago First seen · 40 lines · 140 tokens per session scan A a62153bf99e3
steer-every is a skill published in the GitHub repository wiber/thetacog-mcp (1 stars, last pushed yesterday), licensed MIT. It adds 140 tokens to every session and 1,078 once invoked, about $0.0006 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 2026-10-30.
steer-every compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| steer-everywiber/thetacog-mcp | Skill | 1 | yesterday | MIT | A | 140 | Claude Code, Codex |
| computer-usenexus-research-lab/nexus | Skill | 160 | 2d ago | Apache-2.0 | A | 51 | Claude Code, Codex |
| markedit-toolssammcj/agentic-coding | Skill | 169 | 3d ago | Apache-2.0 | A | 16 | Claude Code, Codex |
| automationnexus-research-lab/nexus | Skill | 160 | 2d ago | Apache-2.0 | A | 76 | Claude Code, Codex |
| deferred-task-executionsammcj/agentic-coding | Skill | 169 | 3d ago | Apache-2.0 | A | 54 | Claude Code |
| ap-cronagentproto/ts | Skill | 9 | today | Apache-2.0 | A | 70 | Claude Code, Codex |
| pb-supervise-parallel-missionagentproto/ts | Skill | 9 | today | Apache-2.0 | A | 58 | Claude Code, Codex |