autoskill

Observe the user's screen via screenpipe, detect repeated research workflows, match them against existing scientific-agent-skills, and draft new skills (or composition recipes that chain existing ones) for the patterns not yet covered. Use when the user asks to analyze their recent work and propose skills based on what they actually do. Requires the screenpipe daemon (https://github.com/screenpipe/screenpipe) running locally on port 3030 — the skill has no other data source and will refuse to run if screenpipe is unreachable. All detection runs locally; only redacted cluster summaries reach the LLM.

Quality

87%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Medium

Suggest reviewing before use

Quality

Content

88%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 well-engineered: executable commands, a validated multi-step workflow with preflight and dry-run gates, and clean one-level-deep references to real bundle files. Its only mild weakness is a few paragraphs that restate the description's privacy/network points, adding modest token overhead.

Dimension	Reasoning	Score
Conciseness	The body is dense and mostly efficient — architecture diagrams, inline config snippets, and numbered pipeline steps earn their tokens — but a few spots over-explain (e.g. the privacy/network blockquote and the 'Composition with other skills' section partly restate the description), keeping it just shy of lean.	4 / 5
Actionability	Copy-paste-ready commands abound — `python scripts/autoskill.py doctor/run/promote` invocations, `pipenv install ...`, `lms load ...`, `export SCREENPIPE_TOKEN=$(screenpipe auth token)` — and the internal pipeline enumerates concrete functions with default thresholds; the common cases are fully covered.	5 / 5
Workflow Clarity	A clear numbered pipeline (Fetch→Redact→Cluster→Match→Synthesize→Report) with a `doctor` preflight checkpoint, a `--dry-run`/`--plan` inspection gate, and a `promote` step that refuses to overwrite existing skills — explicit validation steps and a feedback loop for destructive/batch promotion.	5 / 5
Progressive Disclosure	The SKILL.md is an overview that signals one-level-deep references to real bundle files (`references/screenpipe-config.yaml`, `references/https-proxy.md`) and the `scripts/` directory with a per-file architecture map; no nested/deep references and the config detail lives in `config.yaml` rather than inlined.	5 / 5
	Total	19 / 20 Passed

Description

86%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 strong: it concretely states multiple capabilities, supplies explicit 'Use when' trigger guidance, and carves out a distinct, conflict-resistant niche via its hard screenpipe dependency. It could nudge from 4 toward 5 by echoing the exact verbatim trigger phrases users would naturally say.

Suggestions

Mirror the literal trigger phrases from the 'When to Use' section (e.g. 'Analyze my last 4 hours', 'find composition recipes for workflows I repeat') so the description scores the most natural user-spoken terms.

Add one concrete artifact keyword (e.g. 'proposes new SKILL.md drafts') so the deliverable is explicit in the description itself.

Dimension	Reasoning	Score
Specificity	Quotes 'detect repeated research workflows', 'match them against existing scientific-agent-skills', and 'draft new skills (or composition recipes that chain existing ones)' — multiple concrete actions named with minor gaps (e.g., clustering/matching mechanics are only summarized).	4 / 5
Completeness	Clearly answers 'what' (observe/detect/match/draft via screenpipe) and 'when' ('Use when the user asks to analyze their recent work and propose skills based on what they actually do'), with concrete trigger guidance and an explicit prerequisite/refusal condition.	5 / 5
Trigger Term Quality	Natural trigger phrases like 'analyze their recent work and propose skills' and 'composition recipes that chain existing ones' are present, but the description leans on paraphrased phrasing rather than the exact verbatim phrases ('Analyze my last 4 hours') that users would say, missing some natural variations.	4 / 5
Distinctiveness Conflict Risk	The screenpipe-dependent workflow-mining niche is highly specific and clearly bounded by the data-source prerequisite and refusal clause, making overlap with other skills minimal.	5 / 5
	Total	18 / 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
metadata_field	'metadata' should map string keys to string values	Warning

	Total	15 / 16 Passed

Repository: K-Dense-AI/scientific-agent-skills
Path: skills/autoskill/SKILL.md
Commit: b2a92ba

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.