How do I install the Converge skill?

Run npx skillmds@latest add agentii-ai/converge in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Converge safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Converge?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Converge free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Converge?

agentii-ai (@agentii-ai) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Converge by agentii-ai — Agent Skill

Name: Converge
Author: agentii-ai

Converge

Append-only gap closure and the cadence engine for research theses. Evaluates artifact CURRENT state (never git history, never task checkboxes) against current pins and wrong_if falsifiers, then appends a Convergence section to tasks.md re-proposing the gaps. A clean run leaves tasks.md byte-identical. Deterministic gaps (stale, invalidated) are uncapped; judgment-class gaps are capped per spec 046 Q40.

agentii-ai be24bad 2 files · 2.8 KB Updated Sep 17, 2026

File contents

agentii.converge

Append-only gap closure — the cadence engine of the research orchestration system (spec 046 Q26).

When to run

After every agentii.implement batch (the system detects its own staleness).
On any cadence cycle (quarterly re-visits, post-earnings, post-constitution-bump).
Before any IC review (the conflict list must be current).

What it does

Evaluates current state — artifacts on disk, pins in frontmatter, wrong_if falsifiers in thesis.md. It never reads git history and never trusts [x] checkboxes (Q26 A+: the ledger may lie; the correction is deterministic and has a bounded expiry — the next converge).
Appends, never rewrites — the only write is a ## Phase N: Convergence section at the end of tasks.md, with src: refs and content-derived finding IDs. Existing lines, including previous Convergence sections, are untouched.
Clean run = byte-identical — no gaps means no empty section header.

Gap types

Type	Class	Cap
`stale` (pins older than current)	deterministic	uncapped
`invalidated` (`wrong_if` metric+threshold triggered)	deterministic	uncapped
`missing` / `partial` / `contradicts` / `unrequested`	judgment	Q40 cap (≈50)

Invocation

python3 scripts/converge.py --thesis theses/001-ai-semiconductors \
  --pins '{"assumption_pin": 1, "corpus_version": "2026-08", "as_of": "2026-09-08", "constitution_pin": "0.1.0", "skill_pin": "recent-quarter:abc"}'

Hard rules (Q26)

Never modify spec.md or plan.md.
Never rewrite, renumber, reorder, or delete any existing task.
Finding IDs are content-derived (hash(entity+metric+period+gap_type)) — an unchanged re-run appends nothing.

---name: convergedescription: Append-only gap closure and the cadence engine for research theses. Evaluates artifact CURRENT state (never git history, never task checkboxes) against current pins and wrong_if falsifiers, then appends a Convergence section to tasks.md re-proposing the gaps. A clean run leaves tasks.md byte-identical. Deterministic gaps (stale, invalidated) are uncapped; judgment-class gaps are capped per spec 046 Q40.---# agentii.convergeAppend-only gap closure — the cadence engine of the research orchestration system (spec 046 Q26).## When to run- After every `agentii.implement` batch (the system detects its own staleness).- On any cadence cycle (quarterly re-visits, post-earnings, post-constitution-bump).- Before any IC review (the conflict list must be current).## What it does1. **Evaluates current state** — artifacts on disk, pins in frontmatter, `wrong_if` falsifiers in `thesis.md`. It never reads git history and never trusts `[x]` checkboxes (Q26 A+: the ledger may lie; the correction is deterministic and has a bounded expiry — the next converge).2. **Appends, never rewrites** — the only write is a `## Phase N: Convergence` section at the end of `tasks.md`, with `src:` refs and content-derived finding IDs. Existing lines, including previous Convergence sections, are untouched.3. **Clean run = byte-identical** — no gaps means no empty section header.## Gap types| Type | Class | Cap ||---|---|---|| `stale` (pins older than current) | deterministic | uncapped || `invalidated` (`wrong_if` metric+threshold triggered) | deterministic | uncapped || `missing` / `partial` / `contradicts` / `unrequested` | judgment | Q40 cap (≈50) |## Invocation```bashpython3 scripts/converge.py --thesis theses/001-ai-semiconductors \ --pins '{"assumption_pin": 1, "corpus_version": "2026-08", "as_of": "2026-09-08", "constitution_pin": "0.1.0", "skill_pin": "recent-quarter:abc"}'```## Hard rules (Q26)- Never modify `spec.md` or `plan.md`.- Never rewrite, renumber, reorder, or delete any existing task.- Finding IDs are content-derived (`hash(entity+metric+period+gap_type)`) — an unchanged re-run appends nothing.

Frequently asked questions