phase-gate

phase-gate is a skill for Claude Code from andreymudri/claude-teammates. It costs 31 tokens per session (5,493 once invoked), scanned A, original, MIT.

A verification workflow that checks whether a completed phase of a multi-agent coding run is ready to integrate. It runs command checks and coordinates additional agent or MCP checks, producing a pass or fail result.

In plain words
What is it for?
Running phase verification, executing pending reviewer checks, diagnosing failed gates, and deciding whether a phase can proceed to integration.
Why use it?
It creates a defined quality checkpoint between implementation phases and integration. It also prevents malformed or silently invented checklists from being treated as valid verdicts.

Skill for Claude Code

Written for Claude Code: $CLAUDE_PLUGIN_ROOT variable.

Runs only inside its plugin — its command needs a path that Claude Code sets for a plugin’s own hooks and for nothing else. Install the plugin, not this.

Part of the claude-teammates plugin — 14 skills, 3 agents, 2 hooks shipped together

Good fit Running phase verification, executing pending reviewer checks, diagnosing failed gates, and deciding whether a phase can proceed to integration.

Compare 6 skills from other repositories ↓
Install

Getting it into your agent

This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.

Claude Code
/plugin marketplace add andreymudri/claude-teammates
Claude Code
/plugin install claude-teammates

Made for: Claude Code.

Or install claude-teammates, the plugin that ships this one along with the rest of its 14 skills, 3 agents, 2 hooks.

Wrote 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.

agentmods badge for phase-gate

README.md
[![agentmods](https://agentmods.dev/badge/skills/andreymudri/claude-teammates/phase-gate/github.svg)](https://agentmods.dev/skills/andreymudri/claude-teammates/phase-gate)
Your own site
<a href="https://agentmods.dev/skills/andreymudri/claude-teammates/phase-gate"><img src="https://agentmods.dev/badge/skills/andreymudri/claude-teammates/phase-gate/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.

agentmods 80×15 button for phase-gate

Your own site · 80×15
<a href="https://agentmods.dev/skills/andreymudri/claude-teammates/phase-gate"><img src="https://agentmods.dev/badge/skills/andreymudri/claude-teammates/phase-gate.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 31 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 5,493 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00031 $0.05493
Opus 5 $0.00015 $0.02746
Sonnet 5 $0.00006 $0.01099
Haiku 4.5 $0.00003 $0.00549

Measured 9d ago against content hash 6bef7cb995c4, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

phase-gate 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 9d 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.

skills/phase-gate/SKILL.md · 337 lines

How it starts

The opening of the file, as written. The whole thing — 337 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Phase Gate

Phases run autonomously end to end. The boundary is where verification happens.

Run it

node "$CLAUDE_PLUGIN_ROOT/scripts/cli.mjs" gate --run <runId> --root <project root> [--phase <name>]

Exit codes: 0 PASS, 1 FAIL, 2 the manifest is broken, 3 no manifest (an inferred one was printed).

On exit 3, show the user the inferred manifest, get confirmation, and save it as teammates.gate.json. Never invent checks silently.

Exit 2 is not a verdict — nothing was judged. teammates.gate.json is present and malformed, and the message names the file and what is wrong with it. Show that message and stop; do not re-run the gate, and never save an inferred manifest over the broken one, which would discard the checks the operator meant to fix. complete and fix read the same manifest and exit 2 the same way.

Finish the pending checks

The CLI runs command checks itself and returns agent and mcp checks as pending — those you execute:

  • agent — generate the dispatches rather than assembling them by hand:

    node "$CLAUDE_PLUGIN_ROOT/scripts/cli.mjs" review-dispatch --run --root --phase [--models ]

    It prints one reviewer per lens with the tier, effort, findings path, scratch worktree and prompt already resolved, and exits 4 rather than emitting a dispatch when the phase has no task branch to review. Every rule below is what it encodes; they are stated here because a dispatch assembled by hand still has to follow them.

    Dispatch one tm-reviewer per lens in parallel over the phase diff, without a name. A named reviewer becomes an addressable teammate that goes idle without emitting its result, and the review is lost; unnamed reviewers return normally. Six consecutive named dispatches were lost this way in run preview before the pattern was identified. Each dispatch carries the configured reviewer tier and effort, read with config get agents.reviewer.tier and config get agents.reviewer.effort. config get on an unset key exits 2 with unset: <key> — the same exit code every hard config failure uses, but here it is the normal case, not an error. The two keys fall back differently, and treating them alike is how a reviewer ends up judging below its guaranteed tier:

    • unset: agents.reviewer.tier — dispatch at the fixed reviewer tier, capable (model opus, per the tier→model map in parallel-execution — remap there and this line follows). Never omit the model to inherit the session's: in a mid session that would have the reviewer grading every agent check a full tier below what this skill guarantees. A configured tier replaces capable; nothing else does.
    • unset: agents.reviewer.effort — omit the effort option, and the dispatch inherits the session's effort. Only effort falls back this way.

Read the full file on GitHub · 337 lines

Changes

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.

  1. 9d ago First seen · 337 lines · 31 tokens per session scan A 6bef7cb995c4

Subscribe to this mod's changes

phase-gate is a skill published in the GitHub repository andreymudri/claude-teammates (2 stars, last pushed 7d ago), licensed MIT. It adds 31 tokens to every session and 5,493 once invoked, about $0.0002 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 2026-08-31.