Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add V-Songbird/slag/plugin install colletWrote 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/v-songbird/slag/task-harness)<a href="https://agentmods.dev/skills/v-songbird/slag/task-harness"><img src="https://agentmods.dev/badge/skills/v-songbird/slag/task-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/v-songbird/slag/task-harness"><img src="https://agentmods.dev/badge/skills/v-songbird/slag/task-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 | $0.00142 | $0.03445 |
| Opus 5.5 | $0.00057 | $0.01378 |
| Sonnet 5.5 | $0.00028 | $0.00689 |
| Haiku 4.5 | $0.00014 | $0.00345 |
Grade A, and why
task-harness 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 5d 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 — 233 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Task harness
The whole design is one sentence: prose is read, a hook is executed, and a check that has never been shown to fire is not a check. Everything below follows from that.
Read the plugin's README.md — two directories up — for what ends up in the project. This file is how to mount it,
and the mounting itself is a command — scripts/mount.mjs — because the part that must be
identical every time has no business being a list of instructions someone follows by hand.
Before you write anything
Ask. Do not infer. Read the repository first — its README, its test command, its decisions docs, its git log — so the questions you ask are only the ones the repository cannot answer:
- What is this project, in one line, and what does it run on?
- What is the first task, and what command proves it works?
- What must never happen here without asking? (deploy, spend, publish, delete, install)
- Do you want checks that stop tests being switched off or checks being weakened during a task?
- Which conventions must a change here respect, one sentence each?
Every answer comes from this project's own files or from the person. Never read outside the
project root to find one, sibling projects included: another project's .collet/config.json
describes that project, not this one.
For question four, offer the bundle for the languages the project uses: JavaScript/TypeScript,
Python, Go, Rust, JVM or .NET. From the corresponding catalogue/<id>.json, extract only title,
example and gapNotes for this conversation; the fixture bodies belong to the runner.
Show one short example per mistake and one plain limit drawn from its gapNotes. Keep
the internal ids out of the conversation. The choices are the whole bundle or none for now;
another mistake can be added later with check-writer.
The bundle is optional. Include it when the person asks for it or accepts that choice; do not ask again when their request already covers it. These are source-pattern checks; they do not install compilers, linters or language runtimes. Do not offer coverage for languages outside that list.
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.
- 5d ago Changed · +1 lines 3bf6a4dde798
- 9d ago Changed · +11 lines f3f8ff99201f
- 13d ago Changed 16e8bcbe2f75
- 20d ago Changed · +17 lines 95d4fc6b9e3d
- 22d ago Changed · +29 lines b1d342f371a4
- 23d ago First seen · 175 lines · 142 tokens per session scan A 5687b7d9d417
task-harness is a skill published in the GitHub repository V-Songbird/slag (1 stars, last pushed 6d ago), licensed MIT. It adds 142 tokens to every session and 3,445 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 2026-09-23.
Other skills, from other repositories
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
eval-dataset-design
A guide to creating reliable test sets for AI agents. Each evaluation task includes the request, the simulated user's behaviour, a reset starting state, and a success check that can be independently verified.
tika-eval-compare
Compare extracts from two Tika builds over a corpus to detect regressions in content, encoding, exceptions, and embedded-document handling. Use for "compare before/after extracts", "eval this change against the corpus".
cua-driver
Use Cua Driver for desktop or browser tasks that are awkward or unavailable through Bash/APIs, or when the user explicitly wants GUI interaction: app testing, visual bug reproduction, form filling, calendar entry, screenshots, and demo recording. Also covers Cua setup; not OpenAI Codex Computer Use or web research.
jetson-validate-image
Use after jetson-flash-image to run static BSP checks, on-target smoke/regression tests on a flashed DUT, or both. Not for build or flash steps. Triggers: validate bsp, on-target validation.
atmos-validation
Validate Atmos projects, components, arbitrary JSON Schema inputs, EditorConfig, and GitHub Actions; use affected-file selection and native CI annotations.