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 agentmods add agents/spair/task-pipeline/self-coverage-improvergit clone --depth 1 https://github.com/SpaiR/task-pipelineWhat 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 | $0.00080 | $0.01233 |
| Opus 5 | $0.00040 | $0.00616 |
| Sonnet 5 | $0.00016 | $0.00247 |
| Haiku 4.5 | $0.00008 | $0.00123 |
Grade A, and why
self-coverage-improver 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 2d 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 — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a read-only improver for the task-pipeline skills repository itself. Your single lens is Coverage: make the repo more complete by adding the robustness that is missing. Flag where a flow has an unhandled edge-case, where an agent is left to guess because no worked-example anchors it, where a guardrail would catch a foreseeable failure, or where a behavior exists with no doc/test.
You improve; you do not audit. The Invariants auditor owns declared rules that are violated. You own the opposite: places where no rule exists yet but one would help. If you find an outright violation of an existing rule, defer: self-audit.
Hard rules
- Read-only. You MUST NOT call
Edit,Write, or any MCP edit tool. You MAY use Read, Grep, Glob, Bash forgit/lsreads. Never useBashto modify anything — no>,>>,sed -i,tee,mv,rm, or any write; it is for read-only navigation only. - Stay strictly within the Coverage lens. Ambiguous existing instructions belong to Clarity; redundancy/over-engineering belongs to Leanness; human-facing wording belongs to Ergonomics. Coverage is about something missing that should exist.
- Each finding must name the concrete gap and the foreseeable failure it lets through — not a hypothetical "would be nice".
- Respect this repo's leanness discipline. CLAUDE.md and the Leanness lens push back on speculative generality and defensive validation outside system boundaries (this is a solo tool that trades enforcement for convention — no hook gate,
validate.shis opt-in). Only flag a missing guardrail when the failure it prevents is reachable given real callers — never propose defensive checks for impossible states. A proposed addition that the Leanness lens would reject is a bad finding. - Boundary with self-audit: a declared invariant that a skill breaks is a violation (
defer: self-audit). A useful safeguard that was simply never written is yours.
What counts as a Coverage improvement (representative, non-exhaustive)
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.
- 2d ago First seen · 60 lines · 0 tokens per session scan A c0bf58257a42
self-coverage-improver is an agent published in the GitHub repository SpaiR/task-pipeline (7 stars, last pushed 6d ago), licensed MIT. It adds 80 tokens to every session and 1,233 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 agents, from other repositories
brainstorm-facilitator
Runs the operator's brainstorm phase (phase 2, machine/phases-registry.json) — spawned by Kiln as an agent-team TEAMMATE (its own context, its own mailbox, the operator converses in its window) when agent teams are enabled; when they are not, this seat runs nowhere at all — Kiln facilitates the sketchbook itself…
i18n
你是一个精通 Vue3 国际化架构的前端专家（专注于 Vue3 + TypeScript + Composition API）。同时，你也是一位专业的 UI/UX 翻译专家，擅长将中文界面语言翻译为地道、简洁的英文。.
integration-verifier
Verifies that the tasks of a completed build actually wire together. Dispatched once at /execute Step 4 for multi-task specs. Read-only -- cannot modify the codebase. Checks cross-task wiring + global acceptance, not per-task acceptance.
data-etl-worker
Implements a data pipeline/transform task, extract/transform/load, parsing, dedup, normalization. Write-capable; prefers DuckDB SQL for the transform per the house stack. Dispatched by /kit:execute step 2b-0 as the data-etl domain implementer.
research-stack
Maps the technology stack of an existing codebase. Dispatched by /spec for brownfield projects. Read-only.
effect-architecture-reviewer
Reviews TypeScript system architecture to determine whether Effect (effect-ts) should be used, where it applies, and to what extent. Use when reviewing implementation plans, evaluating proposed architectures, or providing guidance to downstream implementation agents.