pi-anchorloop

Agent workflow skills for pi: one entry skill per task type, each with playbooks, over a shared Align → Anchor → Act → Verify loop, plus loop extensions (loop book, anchor gate, decision trail tool, Jev second thought, skill validator, stream rules).

Packages

Package details

extensionskill

Install pi-anchorloop from npm and Pi will load the resources declared by the package manifest.

npm repo home report

$ pi install npm:pi-anchorloop

Package: pi-anchorloop
Version: 2.16.2
Published: Nov 13, 2026
Downloads: 7,530/mo · 3,674/wk
Author: weihung1017
License: MIT
Types: extension, skill
Size: 466.9 KB
Dependencies: 0 dependencies · 1 peer

Pi manifest JSON

{
  "skills": [
    "./skills"
  ],
  "extensions": [
    "./pi/validate-tool-use.ts",
    "./pi/ttsr.ts",
    "./pi/book.ts",
    "./pi/anchor-gate.ts",
    "./pi/decision-tool.ts",
    "./pi/second-thought.ts"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

pi-anchorloop

繁體中文

Agent workflow skills and loop extensions for pi: every task enters through its own skill, picks a playbook, and runs one shared loop, Align → Anchor → Act → Verify.

Documentation in English and Traditional Chinese: concepts, skills, merge trains, and every extension behavior. It also covers using anchorloop with pi-straw-boss.

Why

An anchor before the first change. Each task names the exact check that observes its result, then acts in small increments and reports what the check saw.
A book that reads evidence. The loop's phase comes from edits, check results, reviews, and landings, not from what the agent says.
Friction becomes rules. An instruction that misled a run is recorded and fixed at its root by solid-loop.

Skills

anchorloop            core loop · principles · playbooks: autonomous-run, pause, pickup
├─ answering        explain · compare
├─ investigating    research · diagnose
├─ inspecting       diff-review · qa-release
├─ coding           bug-fix · feature · refactoring · perf · prototype · state-model  (+ Mini SDD)
├─ integrating      receive-car · train · urgent-car  (+ seams)
├─ deploying        deploy · rollback
├─ writing          draft · tighten · comments · pr-description
├─ visual-designing compare · refine · motion  (+ canvases, measure script)
└─ solid-loop       apply-friction · author-skill · bootstrap
shared              grilling · domain-modeling · human-feedback (uat) · agent-terse

Extensions

File	What it does
`pi/book.ts`	The loop book: opens loops, reads evidence, computes phases, shows the status widget, tracks units and landings, resumes a handed-off book.
`pi/anchor-gate.ts`	Warns once when an edit starts the act phase before an earlier phase is complete.
`pi/decision-tool.ts`	The `anchorloop_decision` tool: a decision trail TSV for runs the user reviews later.
`pi/second-thought.ts`	Asks Jev to check the phase machine's reading when a run settles; advisory only.
`pi/validate-tool-use.ts`	Validates skills, instruction files, and specs after each edit.
`pi/ttsr.ts`	Time Traveling Stream Rules: a rule fires when the agent's own output matches it.

Install

pi install npm:pi-anchorloop
pi install npm:pi-better-show   # optional: review pages for human-feedback and visual-designing

Quick start

Give pi a task; it picks the entry skill whose description matches.
Read the opening Alignment: line and correct a misreading right away.
The first todo opens a loop; the widget below the editor shows its phase, and anchorloop_stage read shows it in full.
Read the reply: it carries what the anchor observed, and anything it did not exercise is unknown.

Develop

uv run python hooks/validate_all.py .   # skills, rules, and durable specs
uv run --with pytest pytest -q          # validators and the pi extensions (needs node)
cd site && npm ci && npm run check && npm run build   # the documentation site

Acknowledgments

pstack by Lauren Tan (MIT), mirror revision e44668c of cursor/plugins (the 2.8.0 engineering-standard additions from cursor/plugins c742dd6), and its port pstack-claude by Michael Denyer (MIT): task-type playbooks, named principles, the lifecycle playbooks, the decision trail, the closed review gate, fresh-subagent hand-off, and evidence grades; the state-model playbook adapts the statechart skill (cursor/plugins a5539b15).
google-research/harness2 (Apache-2.0), revision 1cfe254: solid-loop's placement test, cheapest-mechanism order, and declining a lesson without a before/after contrast.
mattpocock/skills (MIT): grilling, domain-modeling, the diff review, the prototype playbook, and the feature playbook's interface checks.
JuliusBrussee/caveman (Apache-2.0), revision bad48a0: the terse voice in agent-terse.
HumanLayer's show-me (MIT): the views in answering.
oh-my-pi: Time Traveling Stream Rules.

Ported files credit their sources in their own footers.