Name: shepherd-driver
Rating: 73.6 (1 reviews)
Author: microsoft

shepherd-driver

Use only as the composed drive-to-merge stage of an APM batch orchestrator (batch-bug-shepherd, apm-issue-autopilot) that has already selected ONE open pull request in microsoft/apm. Do NOT use for user-facing requests to triage issues, sweep a queue, or open PRs -- the parent orchestrator owns those. Spawn one shepherd-driver subagent per PR: it classifies copilot-pull-request-reviewer[bot] inline review, runs the apm-review-panel, folds (by default) every recommendation inside the PR's stated scope, pushes to the head branch or a superseding PR that preserves authorship via commit trailers, watches CI to green, and iterates under fixed caps until ready-to-merge, advisory-with-deferred, superseded, or blocked. Also provides the cross-PR conflict-resolution and mergeability-gate phase. This is NOT a standalone entrypoint.

Quality

92%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

85%Weight 40%Scale 1-3

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured, highly actionable overview with a clear phased workflow, explicit validation gates, and clean one-level-deep progressive disclosure to real bundle files. The only weakness is duplicated owner-gate/dual-guardrail text across two sections.

Suggestions

De-duplicate the canonical-owner gate and dual-guardrail text: state it once (e.g. in the Disciplines section) and have Phase X.2.5 cross-reference it rather than repeating the full paragraph.

Consider moving the long Phase X.2.5 owner-gate detail into references/ (alongside mergeability-gate.md) and keeping SKILL.md to a one-line pointer, reducing body length without losing the workflow sequence.

Dimension	Reasoning	Score
Conciseness	Mostly lean and assumes Claude's competence (no explaining what a PR/CI is), but the canonical-owner gate and full dual-guardrail text are repeated almost verbatim in both Phase X.2.5 and the Disciplines section -- genuine tighten-able redundancy rather than the every-token-earns-its-place of a 3.	2 / 3
Actionability	Concrete and executable throughout: a copy-paste probe shell snippet, exact CLI commands ('uv run --extra dev ruff check src/ tests/', 'ruff format --check', gh/git), named scripts with --help, and exact phase/cap values; not 2 because there is real executable code rather than pseudocode.	3 / 3
Workflow Clarity	The convergence loop is an explicitly sequenced X.0-X.8 phase list with validation checkpoints (owner gate FAIL CLOSED, lint-until-silent, CI-observed-green, mutation-break gate) and error-recovery feedback loops (cap-hit -> advisory-with-deferred, CI cap -> blocked, malformed return -> re-spawn once -> blocked).	3 / 3
Progressive Disclosure	SKILL.md is a clear overview with one-level-deep, well-signaled references -- all 9 referenced bundle files exist, each is annotated with its phase/purpose in a Bundled assets section, and content is appropriately split (overview in SKILL.md, spawn bodies/rubrics/schemas/scripts in the bundle).	3 / 3
	Total	11 / 12 Passed

Description

100%Weight 40%Scale 1-3

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific, third-person description that names concrete actions, gives explicit use/anti-use triggers, and carves out a clear composed-building-block niche. It fully answers what and when for its intended composing audience.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions -- 'classifies copilot-pull-request-reviewer[bot] inline review, runs the apm-review-panel, folds (by default) every recommendation... pushes to the head branch or a superseding PR... watches CI to green, and iterates under fixed caps' -- matching the multiple-specific-actions anchor.	3 / 3
Completeness	Explicitly answers both what (classify/panel/fold/push/watch-CI/iterate) and when via the opening 'Use only as the composed drive-to-merge stage of an APM batch orchestrator... that has already selected ONE open pull request,' plus explicit anti-triggers ('Do NOT use for user-facing requests to triage issues, sweep a queue, or open PRs').	3 / 3
Trigger Term Quality	Good coverage of the natural terms its composing-audience would use -- 'drive-to-merge,' 'APM batch orchestrator,' 'per-PR convergence loop,' 'mergeability-gate,' 'conflict-resolution,' 'CI to green' -- apt and varied, not generic jargon; not 2 because it goes well beyond a single thin phrase.	3 / 3
Distinctiveness Conflict Risk	Highly distinctive niche -- 'APM batch orchestrator,' 'copilot-pull-request-reviewer[bot],' 'mergeability-gate phase' -- with explicit anti-trigger scope, making it unlikely to fire for the wrong skill; not 2 because it is far from a generic overlap case.	3 / 3
	Total	12 / 12 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
relative_links	Relative link issues: 2 suspicious	Warning
referenced_paths_exist	Referenced path issues: 1 missing	Warning

	Total	14 / 16 Passed

Repository: microsoft/apm
Path: .agents/skills/shepherd-driver/SKILL.md
Commit: 634f7b6

Reviewed: 8 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.