CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

research-pipeline

Full end-to-end research pipeline: from a broad research direction through idea discovery, experiments, and review all the way to a polished paper PDF. Use when user says "全流程", "full pipeline", "从找idea到投稿", "end-to-end research", or wants the complete autonomous research lifecycle.

Quality

79%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./skills/research-pipeline/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

66%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.

A well-structured orchestrator skill with excellent sequencing, gates, and resume/acceptance mechanics, but its token budget suffers from meaningful duplication — the heartbeat doctrine and several constant rules each appear twice or more. Unbound shell variables in otherwise copy-paste-ready snippets slightly reduce executability.

Suggestions

Deduplicate the heartbeat/external-cadence material: keep a one-line pointer at the top and retain the full doctrine (resolver chain, pivot rules) in only one place — either the single section or the referenced shared-references/external-cadence.md.

Bind $ROOT, $RUN_ID, and $STAGE where they are first used in the run-state and heartbeat sections so the bash snippets are directly executable.

Collapse the repeated AUTO_PROCEED/VENUE/RENDER_HTML explanations (Constants, Checkpoint execution rule, Gate sections, Key Rules) into one canonical statement, leaving the other mentions as one-line cross-references.

DimensionReasoningScore

Conciseness

The heartbeat/external-cadence doctrine appears nearly wholesale twice (the "⏱ External cadence" blockquote and the "Overnight heartbeat" section, each with its own copy of the helper-resolution chain), and AUTO_PROCEED, VENUE, and RENDER_HTML rules are each restated across Constants, the "Checkpoint execution rule", the gate sections, and "Key Rules". That is several padded/redundant sections rather than a single tightening opportunity, so it sits below the mostly-efficient anchor of 3, though it never drifts into explaining concepts Claude already knows (not 1).

2 / 5

Actionability

Guidance is largely executable: exact sub-skill invocations ('/experiment-bridge "$CHOSEN_IDEA_TITLE" — code review: $CODE_REVIEW, base repo: $BASE_REPO'), concrete state commands ('run_state.py start <root> <run_id> --phases "..."'), and a copy-paste bash resolver chain. It falls short of fully executable because $ROOT, $RUN_ID, and $STAGE are used in the heartbeat/run-state snippets without ever being bound, and several code blocks are templates with unresolved placeholders.

4 / 5

Workflow Clarity

The five stages are clearly sequenced with explicit validation: an acceptance table per phase ("Mark `accepted` ONLY after the stage's gate passes"), cross-model gates, re-validation of done-but-unaccepted stages on resume, and feedback loops (sanity check with "auto-debugs failures (up to 3 attempts)", stop at round 4 without positive assessment, non-blocking render failure handling). This matches the top anchor's explicit checkpoints plus error-recovery loops.

5 / 5

Progressive Disclosure

Detail is appropriately delegated via clearly signaled one-level links (shared-references/reviewer-backends.md, external-cadence.md, resumable-runs.md, the Output Protocols section, templates/RESEARCH_BRIEF_TEMPLATE.md). It is not 5 because the heartbeat doctrine is duplicated inline instead of fully delegated to the referenced file, and the referenced bundle files live outside this skill's directory (no references/ or scripts/ exist here), so navigation depends on unverifiable sibling paths.

4 / 5

Total

15

/

20

Passed

Description

87%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.

A strong description: concrete lifecycle stages, explicit bilingual trigger phrases, and both what-and-when clearly answered. The only soft spot is trigger coverage that omits a few natural single-phrase variants.

DimensionReasoningScore

Specificity

The description lists several concrete pipeline stages — "from a broad research direction through idea discovery, experiments, and review all the way to a polished paper PDF" — which is more than the 1–2 actions of a score-3 anchor. It falls short of 5 because constituent actions (implementation, deployment, writing) are not individually enumerated, leaving minor coverage gaps.

4 / 5

Completeness

It explicitly answers both what ("Full end-to-end research pipeline: ... idea discovery, experiments, and review all the way to a polished paper PDF") and when ("Use when user says ...") with concrete trigger phrases, exactly matching the top anchor. It is not 4 because the 'when' clause is fully explicit, not merely improvable.

5 / 5

Trigger Term Quality

The trigger list "全流程", "full pipeline", "从找idea到投稿", "end-to-end research" gives good natural-phrase coverage including two languages and colloquial forms. A few natural variants are missing (e.g., "research pipeline" alone, "autonomous research"), so it is above the 'some relevant keywords' anchor but below comprehensive.

4 / 5

Distinctiveness Conflict Risk

The "complete autonomous research lifecycle" framing carves out a clear orchestrator niche with distinct multi-stage triggers ("全流程", "从找idea到投稿") that the individual sub-workflows (idea discovery, experiments, review alone) would not match, so conflict risk is minimal.

5 / 5

Total

18

/

20

Passed

Validation

81%

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

Validation — 13 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

allowed_tools_field

'allowed-tools' contains unusual tool name(s)

Warning

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

relative_links

Relative link issues: 8 suspicious

Warning

Total

13

/

16

Passed

Repository

wanshuiyin/Auto-claude-code-research-in-sleep

Path

skills/research-pipeline/SKILL.md

Reviewed

Table of Contents

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.