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 Firstp1ck/pi-coding-agent-forge --skill small-model-regression-evalgit clone --depth 1 https://github.com/Firstp1ck/pi-coding-agent-forgeWrote 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/firstp1ck/pi-coding-agent-forge/small-model-regression-eval)<a href="https://agentmods.dev/skills/firstp1ck/pi-coding-agent-forge/small-model-regression-eval"><img src="https://agentmods.dev/badge/skills/firstp1ck/pi-coding-agent-forge/small-model-regression-eval/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/firstp1ck/pi-coding-agent-forge/small-model-regression-eval"><img src="https://agentmods.dev/badge/skills/firstp1ck/pi-coding-agent-forge/small-model-regression-eval.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 | $0.00040 | $0.00637 |
| Opus 5.5 | $0.00016 | $0.00255 |
| Sonnet 5.5 | $0.00008 | $0.00127 |
| Haiku 4.5 | $0.00004 | $0.00064 |
Grade A, and why
small-model-regression-eval 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 1mo 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 — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Small-model regression evaluation
Use this skill to compare a model, quantization, prompt, guard profile, or skill revision against frozen independent outcomes. It separates deterministic mechanism tests from live model quality results.
When to use
Should trigger
- “Compare this configured local model with the guarded workflow on held-out tasks.”
- “Report whether a reliability revision regressed retrieval, agentic, or coding fixtures.”
- “Create a baseline-versus-integrated evaluation plan without making live calls.”
Should not trigger
- “Implement this feature.”
- “Use a model to solve the current task.”
- “Publish or enable a model.”
Inputs
Provide a frozen task suite, independent expected outcomes, exact provider/model ID, quantization and sampling settings when applicable, package revision, guard profile, tool configuration, timeout, sample budget, and whether assisted mode is separately authorized.
Portable workflow
- Freeze sanitized task inputs and independent success checks before tuning.
- Run deterministic guard fixtures separately from model executions.
- Compare baseline, current extension, isolated feature, integrated guarded, and separately authorized assisted configurations.
- Keep assisted results in their own column; do not combine them with small-model-only scores.
- Record exact model/provider identity and all execution settings for every live run.
- Report unavailable live evaluation truthfully instead of substituting a model, provider, or adapter.
Safety and reporting boundary
Do not use the extension completion gate as the evaluation oracle. Do not make network or provider calls merely because a model name was supplied. Never claim an improvement from fixtures, mechanism tests, or unrun configurations.
Verification
Verify that each actual task outcome is checked by an independent oracle, frozen inputs remain unchanged, configuration identity is complete, failures are retained, and unavailable live execution records zero calls with a reason.
What ships with it
2 files 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.
- 1mo ago First seen · 52 lines · 40 tokens per session scan A afaf16628d3d
small-model-regression-eval is a skill published in the GitHub repository Firstp1ck/pi-coding-agent-forge (91 stars, last pushed today), licensed MIT. It adds 40 tokens to every session and 637 once invoked, about $0.0002 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 2026-09-14.
Other skills, from other repositories
prompt-engineering-expert
Expert prompt engineering, custom instruction, system prompt, and agent instruction design. Use when Codex needs to review, generate, refactor, debug, optimize, document, or test AI prompts; design reusable prompt templates; create or improve system prompts, custom instructions, agent behavior guidelines, tool-use…
prompt-compressor
Compress verbose prompts & context before LLM processing. This skill should be used when input exceeds 1500 tokens, contains redundant phrasing, or includes unnecessary context. Reduces tokens by 40-60%.
gemini-cli
Run Gemini CLI for AI queries. Use when user asks to "run/ask/use gemini", compare Claude vs Gemini, or delegate tasks to Gemini.
prompt-library
Use when looking for a ready-made agent or task template (code review, support, data analysis, translation, extraction) instead of writing one from scratch.
prompt-creation
Use when writing a new system prompt from scratch, structuring it with Anthropic's 9-element format, or choosing emphasis levels and few-shot examples.
prompt-testing
Use when comparing two prompt variants, defining quality/efficiency/robustness metrics, or deciding whether to adopt a challenger prompt over a baseline.