research-deep

Run deep research on each item in outline.yaml by fanning out parallel web-search subagents. Writes one validated JSON per item to results/. Resumable — skips items that already have a complete JSON.

Quality

65%

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

Fix and improve this skill with Tessl

tessl review fix ./plugins/ultra-research/skills/research-deep/SKILL.md

Quality

Content

71%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 content is well-structured and actionable with a clear multi-step workflow, real validation checkpoints, and exemplary progressive disclosure that moves the bulky prompt template into a verified one-level reference. The main weakness is repetitive restatement of the orchestrator hard constraint, which inflates the token budget.

Suggestions

State the orchestrator MUST-NOT-search constraint once in the Hard Constraint block and reference it by name later instead of repeating the full prohibition in Execution Notes and Step 3.

Add an explicit retry/re-dispatch path for items whose <slug>.json never appears after a wave, so silent subagent failure is handled rather than only flagged at summary time.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient and assumes Claude's competence, but the orchestrator MUST-NOT-search constraint is restated four times (table, 'Why this is a hard constraint' prose, Execution Notes, and again in Step 3), which is padding that could be tightened without losing the message.	3 / 5
Actionability	It gives executable bash for validator discovery, concrete glob patterns, and specific tool invocations (Agent/Task with run_in_background), with placeholders bound to real values; minor gaps (no literal subagent launch example inline) keep it just below 5.	4 / 5
Workflow Clarity	The five steps are clearly sequenced with validation checkpoints (resume check in Step 2, subagent-run validator, disk monitoring in Step 4) and a confirm-before-batch gate, satisfying the batch-operation feedback requirement; a small gap in explicit retry-on-silent-failure handling keeps it at 4 rather than 5.	4 / 5
Progressive Disclosure	The body is a clear overview that offloads the verbatim prompt template to references/prompt-templates.md (verified to exist), referenced one level deep with a descriptive bullet listing its contents — exactly the well-signaled single-level reference pattern.	5 / 5
	Total	16 / 20 Passed

Description

58%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 third-person and specific about what the skill does, but it omits any explicit 'when to use' trigger guidance and relies on technical pipeline terms over natural user keywords. This caps completeness and limits trigger-term quality.

Suggestions

Add an explicit 'Use when...' clause naming natural triggers, e.g. 'Use when running the ultra-research pipeline's research phase or when the user asks to research every item in outline.yaml in parallel.'

Soften technical jargon with natural synonyms users would say — 'deep research', 'research every item', 'parallel research' — alongside 'web-search subagents' and 'outline.yaml'.

Dimension	Reasoning	Score
Specificity	Lists several concrete actions — 'fanning out parallel web-search subagents', 'Writes one validated JSON per item to results/', 'skips items that already have a complete JSON' — but coverage is scoped to one pipeline niche rather than comprehensive, so it sits below a 5.	4 / 5
Completeness	The 'what' is clear (research each item, fan-out, write JSON, resume), but there is no 'Use when...' clause or equivalent explicit trigger guidance, which per the rubric caps completeness at 3.	3 / 5
Trigger Term Quality	'deep research' is a natural user phrase, but the rest leans technical ('parallel web-search subagents', 'validated JSON', 'outline.yaml') with no synonyms or common variations a user would actually say, matching the 'some relevant keywords but missing common variations' anchor.	3 / 5
Distinctiveness Conflict Risk	The reference to 'outline.yaml' and the narrow per-item fan-out role give it a distinct niche with minimal conflict risk, though it could mildly overlap with a generic deep-research skill, keeping it just below 5.	4 / 5
	Total	14 / 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
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning

	Total	15 / 16 Passed

Repository: seanGSISG/claude-depot
Path: plugins/ultra-research/skills/research-deep/SKILL.md
Commit: 48bf6a4

Reviewed: 4 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.