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 sky-valley/pi --skill pi-parity-reviewgit clone --depth 1 https://github.com/sky-valley/piWrote 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/sky-valley/pi/pi-parity-review)<a href="https://agentmods.dev/skills/sky-valley/pi/pi-parity-review"><img src="https://agentmods.dev/badge/skills/sky-valley/pi/pi-parity-review/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/sky-valley/pi/pi-parity-review"><img src="https://agentmods.dev/badge/skills/sky-valley/pi/pi-parity-review.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00054 | $0.01448 |
| Opus 5 | $0.00027 | $0.00724 |
| Sonnet 5 | $0.00011 | $0.00290 |
| Haiku 4.5 | $0.00005 | $0.00145 |
Grade A, and why
pi-parity-review 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 11d 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 — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
pi-parity-review — is the port true to pi?
Input: the upstream change (sha in the upstream clone) and our ported diff. You are independent of the porter: assume the port is wrong until the diff proves otherwise. Self-authored tests are circular — they pin what the porter believed, not what pi does. Every sweep of this project found real bugs ONLY via comparison against real pi.
Working-tree safety (hard rule)
You are reviewing the UNCOMMITTED working tree — it is the only copy of the
port. NEVER mutate it: no git checkout/restore/reset/stash/clean, no
edits to ported files. To mutation-test (disable a fix and confirm its test
fails), copy the file to a scratch path, mutate the COPY in a throwaway dir
(or git worktree add a detached checkout), run tests there, and discard it —
never touch the live file. If you ever do dirty the tree, restore it exactly
and say so loudly in your report.
References
- Upstream TS:
$PI_UPSTREAM_DIR(default~/.cache/pi-upstream). Never read its working tree — always name a sha:git -C dir show <sha>:<path>,git -C dir diff <sha>^1..<sha>,git -C dir grep <pattern> <sha> -- <path>. The checkout is whatever was last checked out and is nobody's contract; during the 2026-08-11 cycle it sat at a 2026-06-07 commit, ~2 months behind, so a plaingrepof it reported the OPPOSITE of what the pin contained and read as though upstream had reverted the change under review./pi-syncnow fast-forwards it during preflight, which makes that failure less likely but not impossible — and even a correct tree is atorigin/main, which mid-cycle is AHEAD of the pin and so is still not what you are reviewing. - Published npm build: install/refresh
@earendil-works/pi-ai+pi-coding-agentat the matching release into a scratch dir. When the TS source and the shipped build disagree, the BUILD wins — it's what real pi runs and what all goldens come from. - Goldens in-repo:
coding/testdata/sessparity/,coding/testdata/imgparity/, the default-system-prompt golden, and the tool output-string tests.
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.
- 11d ago First seen · 95 lines · 54 tokens per session scan A e2933de526a8
pi-parity-review is a skill published in the GitHub repository sky-valley/pi (48 stars, last pushed today), licensed MIT. It adds 54 tokens to every session and 1,448 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 2026-08-30.
Other skills, from other repositories
code-review
Run a thorough self-review pass on the most recent change.
release
Release gentle-pi through GitHub and npm. Trigger: release, publish, npm publish, GitHub release, version bump.
gentle-ai
Use Gentle AI harness discipline for Pi work: clarify first, preserve OpenSpec artifacts, use strict TDD where available, delegate through subagents when useful, and protect review workload.
gentle-ai-work-unit-commits
Plan commits as reviewable work units. Trigger: implementation, commit splitting, chained PRs, or keeping tests and docs with code.
gentle-ai-skill-creator
Trigger: /skill-creation, skill creation, skill creator, create skill, new skill. Create LLM-first skills with valid frontmatter.
gentle-ai-judgment-day
Trigger: judgment day, judgement day, dual review, adversarial review, juzgar. Run explicit blind dual review with at most two scoped fix/re-judgment rounds.