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 ShreyPaharia/octomux --skill configure-harnessgit clone --depth 1 https://github.com/ShreyPaharia/octomuxWrote 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/shreypaharia/octomux/configure-harness)<a href="https://agentmods.dev/skills/shreypaharia/octomux/configure-harness"><img src="https://agentmods.dev/badge/skills/shreypaharia/octomux/configure-harness/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/shreypaharia/octomux/configure-harness"><img src="https://agentmods.dev/badge/skills/shreypaharia/octomux/configure-harness.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 | $0.00046 | $0.01780 |
| Opus 5 | $0.00023 | $0.00890 |
| Sonnet 5 | $0.00009 | $0.00356 |
| Haiku 4.5 | $0.00005 | $0.00178 |
Grade C, and why
configure-harness scanned grade C with 2 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 12d 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
-d '{"harnesses":{"claude-code":{"flags":"--foo; rm -rf /"}}}' Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -X PATCH http://127.0.0.1:7777/api/settings \ How it starts
The opening of the file, as written. The whole thing — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Configure the harness octomux launches
octomux launches a harness (a coding-agent CLI) per task/agent. Two ship
built-in — claude-code (default) and cursor — and a plugin can register more.
This skill covers picking one, tuning its flags, and what's still unwired if
you're building against the Harness interface directly.
Per-task model override
The narrowest lever: override the model for one task or one agent, independent of harness settings.
octomux create-task --model claude-opus-4-8 ...
octomux add-agent --task <task-id> --model claude-sonnet-4-6 --prompt '...'
applyModel(flags, model) (server/harnesses/shared.ts) strips any existing
--model <value> out of the resolved flags string and appends the per-task one —
so a global --model in settings.harnesses.claude-code.flags never wins over a
task's own override.
Harness selection
Per-task, at creation time:
octomux create-task --harness-id cursor ...
Omit --harness-id over the CLI/API and task.create hardcodes the literal
'claude-code' (server/registry/capabilities/task.ts) — this does not
consult settings.defaultHarnessId. settings.defaultHarnessId only drives the
dashboard's create-task dialog: HarnessPicker.tsx reads it client-side to
pre-select a value in the dropdown before the user ever submits, so it's a UI
convenience, not a server-side default. Set it from the Settings page in the
dashboard, or:
curl -X PATCH http://127.0.0.1:7777/api/settings \
-H 'Content-Type: application/json' \
-d '{"defaultHarnessId": "cursor"}'
There's no CLI wrapper for /api/settings — use curl against the running
server, or the Settings page in the dashboard. If you want every task to actually
launch on a non-claude-code harness regardless of client, pass --harness-id
explicitly every time (or wrap create-task in your own script/alias) — there is
currently no server-side knob for that. getHarness(id) throws Unknown harness: <id> if nothing registered that id, so a typo in --harness-id surfaces
immediately at task-launch time.
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.
- 12d ago First seen · 174 lines · 46 tokens per session scan C e0d01372b476
configure-harness is a skill published in the GitHub repository ShreyPaharia/octomux (22 stars, last pushed 11d ago), licensed MIT. It adds 46 tokens to every session and 1,780 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
claw-metagenomics
Shotgun metagenomics profiling — taxonomy, resistome, and functional pathways.
clawrouter
Smart LLM router — save 67% on inference costs. Routes every request to the cheapest capable model across 41 models from OpenAI, Anthropic, Google, DeepSeek, and xAI.
company-analyzer
A company-analysis tool that calls language models to produce analyses such as SWOT, valuation, and industry reviews. It includes shell helpers for caching results, tracking API costs, and checking a daily budget.
prompt-tuning
Tune a prompt, or anything whose quality is measured by non-deterministic model output, without chasing noise - a noise baseline before the first edit, medians over repeated runs, enforcement AFTER generation rather than in the wording. Use when iterating on prompts or model-judged output.
memory-patterns
Persistent memory patterns for moflo agents — session memory, long-term knowledge, pattern learning, and cross-session context via moflo's node:sqlite + HNSW vector store. Use when building stateful agents or assistants that need to remember across runs.
vector-search
Semantic vector search with moflo — RAG over your own documents, similarity matching, context-aware retrieval via HNSW (node:sqlite-backed). Use when building retrieval layers for chat, search, or context-assembly.