How do I install the Convergence skill?

Run npx skillmds@latest add yezhwi/convergence 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 Convergence safe to use?

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

Which AI agents work with Convergence?

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 Convergence free to use?

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

Who published Convergence?

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

Convergence by yezhwi — Agent Skill

Name: Convergence
Rating: 4.2 (5 reviews)
Author: yezhwi

File contents

Convergence Skill

Decide whether the harness loop continues. v0.1 uses fixed rules only — no scoring, no model confidence.

Workflow Commands

harness gate preflight
harness impact scope --format yaml
harness evidence run --type unit_test --command "pytest"
harness evidence attach --type build --file external-proof.json
harness finding resume-review FND-001

Rules

PASS

harness gate prints DECISION: CONVERGED
→ harness transition DONE
→ report to user with gate output attached

Never harness transition CONVERGED. Only harness gate may enter CONVERGED.

Continue

All three hold:

Gate BLOCKED
+ 存在明确可处理 blocker（每个 blocker 有明确的下一状态）
+ iteration < max_iterations
→ increment iteration, leave BLOCKED / return to appropriate state

Blocker dispatch:

open finding → REPRODUCING (via reproduce-finding)
unverified requirement / red verification → IMPLEMENTING

Escalate

Any one holds → transition to ESCALATED:

iteration >= max_iterations                          -> detected by harness gate
same finding VERIFIED then open again (regression)   -> detected by harness gate
same invariant repeatedly violated                   -> human/skill declares
test suite unstable                                  -> human/skill declares
architecture defect suspected                        -> human/skill declares
spec ambiguity blocks verification                   -> human/skill declares

Output exactly one reason code:

SPEC_AMBIGUITY | ARCHITECTURE_DEFECT | REPEATED_REGRESSION |
UNSTABLE_TEST  | REVIEW_DISAGREEMENT  | MAX_ITERATIONS

Code-detected codes (MAX_ITERATIONS, REPEATED_REGRESSION) are emitted by harness gate. For the others YOU must declare them explicitly to the user with evidence - never silently retry past a judgment-call blocker.

Hard Boundaries

DONE only via CONVERGED → DONE, and CONVERGED only after DECISION: CONVERGED.
ESCALATED ends the autonomous loop. Report reason + full status to user; do not silently retry.
Never reset iteration to dodge max_iterations.

---name: convergencedescription: Use when `harness gate` has printed DECISION: CONVERGED, CONTINUE, or ESCALATED and the loop must finish, resume, or stop. Applies v0.1 escalation rules with no scoring.---# Convergence SkillDecide whether the harness loop continues. v0.1 uses fixed rules only — noscoring, no model confidence.## Workflow Commands```bashharness gate preflightharness impact scope --format yamlharness evidence run --type unit_test --command "pytest"harness evidence attach --type build --file external-proof.jsonharness finding resume-review FND-001```## Rules### PASS```textharness gate prints DECISION: CONVERGED→ harness transition DONE→ report to user with gate output attached```Never `harness transition CONVERGED`. Only `harness gate` may enter CONVERGED.### ContinueAll three hold:```textGate BLOCKED+ 存在明确可处理 blocker（每个 blocker 有明确的下一状态）+ iteration < max_iterations→ increment iteration, leave BLOCKED / return to appropriate state```Blocker dispatch:- open finding → REPRODUCING (via reproduce-finding)- unverified requirement / red verification → IMPLEMENTING### EscalateAny one holds → transition to ESCALATED:```textiteration >= max_iterations -> detected by harness gatesame finding VERIFIED then open again (regression) -> detected by harness gatesame invariant repeatedly violated -> human/skill declarestest suite unstable -> human/skill declaresarchitecture defect suspected -> human/skill declaresspec ambiguity blocks verification -> human/skill declares```Output exactly one reason code:```textSPEC_AMBIGUITY | ARCHITECTURE_DEFECT | REPEATED_REGRESSION |UNSTABLE_TEST | REVIEW_DISAGREEMENT | MAX_ITERATIONS```Code-detected codes (MAX_ITERATIONS, REPEATED_REGRESSION) are emitted by`harness gate`. For the others YOU must declare them explicitly to theuser with evidence - never silently retry past a judgment-call blocker.## Hard Boundaries1. DONE only via CONVERGED → DONE, and CONVERGED only after `DECISION: CONVERGED`.2. ESCALATED ends the autonomous loop. Report reason + full status to user; do not silently retry.3. Never reset `iteration` to dodge `max_iterations`.

Frequently asked questions

Convergence