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 OutlineDriven/odin-claude-plugin --skill instruction-phrasing-microtestgit clone --depth 1 https://github.com/OutlineDriven/odin-claude-pluginWrote 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/outlinedriven/odin-claude-plugin/instruction-phrasing-microtest)<a href="https://agentmods.dev/skills/outlinedriven/odin-claude-plugin/instruction-phrasing-microtest"><img src="https://agentmods.dev/badge/skills/outlinedriven/odin-claude-plugin/instruction-phrasing-microtest.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00059 | $0.01283 |
| Opus 5 | $0.00030 | $0.00642 |
| Sonnet 5 | $0.00012 | $0.00257 |
| Haiku 4.5 | $0.00006 | $0.00128 |
Grade A, and why
instruction-phrasing-microtest 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 3d 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 — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Instruction phrasing microtest
Contract
| Field | Bound contract |
|---|---|
| Trigger | About to change the wording of a rule in a skill, prompt template, or agent instruction, where the change is meant to alter what the model produces. |
| Authority | Reversible-local: writes only a local results table file; no repository, VCS, credential, paid, published, deployed, or remote mutation. Rollback is deleting the results file. |
| Side effect | Runs a bounded set of cheap single-call model samples against variant system prompts and writes a scored results table. No repository mutation. |
| Done | A results table exists comparing each variant against the control on programmatic markers, with every match hand-verified, and the adopted phrasing beats the control on the target metric without regressing the others — or the change is dropped as unmeasurable. |
Inputs
- Control text (required): the current wording of the rule or instruction being evaluated.
- Variant texts (required): one or more proposed rephrasings. Each variant must be a complete, self-contained instruction — not a diff or delta from the control.
- Fixture scenario (required): a realistic mid-workflow user message or task prompt that would trigger the rule. Must be specific enough to tempt the failure the instruction targets.
- Scoring markers (required): regex patterns or literal strings to search for in model output that indicate compliance or violation. At least one compliance marker and one violation marker.
- Target metric (required): which scoring marker the variant must improve over the control.
- Sample count (optional, default 5): number of model calls per variant. Minimum 5 for statistical signal.
- Model (optional): the model to sample. Must match the model that writes the artifact in production.
Procedure
- Classify the instruction. Determine which category the current rule falls into:
- Tripwire: phrase-level self-check on concrete tokens (e.g., "if your output contains 'do not flag' … stop").
- Recognition table: red-flags or rationalization table read at decision time.
- Discrete-directive prohibition: "Do not ask X to do Y" where the model has no competing incentive to do Y.
- Composition prohibition: a prohibition on how to compose output where the model has its own agenda for the output (e.g., restating specs feels like helpful curation). This classification determines which phrasing strategies are worth testing. Composition prohibitions are the category most likely to backfire; tripwires and recognition tables are the most reliable.
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.
- 3d ago First seen · 73 lines · 59 tokens per session scan A 9d2923fddb79
instruction-phrasing-microtest is a skill published in the GitHub repository OutlineDriven/odin-claude-plugin (35 stars, last pushed today), licensed Apache-2.0. It adds 59 tokens to every session and 1,283 once invoked, about $0.0003 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 2026-09-04.
Other skills, from other repositories
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
local-ai-agents
Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…
next-cache-components-adoption
Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…
next-partial-prefetching-adoption
Turn on Partial Prefetching in a Next.js app and work through the insights it surfaces. Use when the user wants to enable or adopt Partial Prefetching, flip the partialPrefetching flag, opt routes in with export const prefetch = 'partial', audit Link prefetch={true} behavior, preserve existing prefetched UI with…
chronicle
Analyze Copilot session history for standup reports, usage tips, session search, and session reindexing. Use when the user asks for a standup, daily summary, usage tips, workflow recommendations, wants to search or find past sessions by keyword/file/PR, wants to reindex their session store, or asks about deleting…
babysit-pr
Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…