agent-loop

Goal-driven loop harness. Subcommands — init (build the 7-file harness in this repo), new (interview → goal spec), run (plan→act→verify iterations; plans await human approval), approve/reject (act on a pending plan), status (loop health), doctor (harness health), watch (live dashboard), canonize (promote memory to vault). Slash-only.

Quality

72%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./plugins/agent-loop/skills/agent-loop/SKILL.md

Quality

Content

81%Weight 40%Scale 1-5

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

The body is an exceptionally actionable and well-sequenced orchestration manual with validation gates and feedback loops at every stage. Its sole notable weakness is conciseness — emphatic repetition and long asides inflate the token budget beyond what the operational complexity strictly requires.

Suggestions

Trim emphatic redundancies ('never race a live lease', 'the tree is evidence', 'never hand-roll JSON edits') to single statements; the determinism is already enforced by the bin tools.

Move the deeper run-iteration detail (lease/stall/interruption recovery, audit-shape specifics) into references/loop-contract.md and keep the body to the gate-level sequence.

Collapse the long parenthetical asides in PLAN/RECORD into short pointer lines referencing loop-contract.md sections.

Dimension	Reasoning	Score
Conciseness	Mostly substantive operational instruction rather than concept padding, but it carries noticeable emphatic redundancy (repeated 'never race a live lease', 'the tree is evidence', 'never hand-roll JSON') and long parenthetical asides that could be tightened.	3 / 5
Actionability	Provides exact CLI invocations (al-state plan-propose '<JSON>', al-state tdd-red '<command>', curl -s -m 2 http://127.0.0.1:4177/api/fleet), precise JSON return shapes for agents, and concrete file paths throughout — copy-paste ready.	5 / 5
Workflow Clarity	Each subcommand is a numbered sequence; the run flow is WAKE→PLAN→ACT→VERIFY→RECORD→OPTIMIZE with explicit validation gates (al-goal validate, plan-propose gate, tdd-red refusal, al-verify, record-iter re-verify), feedback loops (revise/re-critique/re-validate), and the D01–D16 doctor checklist.	5 / 5
Progressive Disclosure	Clear per-subcommand section structure with well-signaled one-level references to real files (references/loop-contract.md, references/doctor-checks.md#<id>); the only gap is the lengthy inlined run procedure that partly overlaps the dedicated loop-contract.md.	4 / 5
	Total	17 / 20 Passed

Description

63%Weight 40%Scale 1-5

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

The description is highly specific and distinctive, clearly enumerating the skill's subcommands and their concrete purposes. Its main weakness is the absence of an explicit 'when to use' trigger clause and a reliance on internal jargon over natural user language.

Suggestions

Add a 'Use when...' clause naming natural user triggers (e.g. 'Use when you want to drive a goal to completion through autonomous plan→act→verify iterations').

Soften internal jargon ('harness plane', 'canonize', 'vault') or pair it with plain-language equivalents so users recognise the trigger terms.

Include common synonyms a user might say ('agentic loop', 'iterative coding', 'automated goal runner') to broaden trigger-term coverage.

Dimension	Reasoning	Score
Specificity	Enumerates nine subcommands each paired with a concrete action (e.g. 'init (build the 7-file harness)', 'run (plan→act→verify iterations; plans await human approval)', 'canonize (promote memory to vault)'), giving comprehensive coverage of the skill's surface.	5 / 5
Completeness	The 'what' is clear and detailed, but there is no 'Use when...' clause or equivalent explicit trigger guidance; per the rubric this caps completeness at 3.	3 / 5
Trigger Term Quality	Contains some relevant domain keywords (loop, goal, plan, approve, status, watch, doctor) but leans on internal jargon ('harness', 'canonize', 'vault', 'harness plane') and omits natural user phrasings like 'agentic loop' or 'iterate on a coding task'.	3 / 5
Distinctiveness Conflict Risk	The subcommand taxonomy (init/new/run/approve/reject/status/doctor/watch/canonize) is a distinct niche and the skill is slash-only, leaving only minor overlap risk with generic agent/loop skills.	4 / 5
	Total	15 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: belchman/claude-skills
Path: plugins/agent-loop/skills/agent-loop/SKILL.md
Commit: 77a9e6b

Reviewed: 3 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.