Skills catalogue

The plugin ships about 70 skills, grouped along the EMPN research workflow. Three sequential phases (Position → Evidence → Output) plus three transversal layers (Situation awareness, Quality gates, Support).

The phases aren't strictly sequential — you'll loop. Evidence often reshapes position, and quality gates apply at every step.

flowchart LR
    SA["<b>1. Situation awareness</b><br/><i>transversal</i>"]

    subgraph LOOP["sequential phases — iterate freely"]
        direction LR
        P["<b>2. Position</b><br/><b>formation</b>"]
        E["<b>3. Evidence</b><br/><b>base</b>"]
        O["<b>4. Public-facing</b><br/><b>output</b>"]
        P --> E --> O
        O -.->|iterate| P
        E -.->|reshape| P
    end

    QG["<b>5. Quality gates</b><br/><i>cross-phase</i>"]
    SUP["<b>6. Support</b><br/><i>transversal</i>"]

    SA -->|feeds context| LOOP
    QG -.->|reviews| LOOP
    SUP -.->|scaffolds| LOOP

By intent — what brought you here?

• Need to know what's happening on a file or who's in the room → row 1 (Situation awareness)
• Need to frame a take, stress-test a proposal, or scope a research project → row 2 (Position formation)
• Need a number, a paper, a primary source, or a distilled extract → row 3 (Evidence base)
• Need to write a brief, validate it, or polish a draft → row 4 (Public-facing output)
• Need to verify, attack, or refine an argument or claim → row 5 (Quality gates)
• Need to scaffold a project, course-correct, capture lessons, or build a skill → row 6 (Support)

The full catalogue

For every important phase there's a Do column (produce the artefact) and a Review column (stress-test it). Review gaps flags missing quality-gate skills planned for the roadmap; Roadmap lists future Do skills.

1. Situation awareness (transversal)

Do	Review
eu-policy-research, eu-legislation-tracker, stakeholder-map, eu-consultation-feedback, contacts-extract, contacts-discover, weekly-digest, eu-agenda-sync	—

Review gaps: stakeholder-position-verifier. Roadmap: trilogue tracker · amendments diff · coalition / co-signatory mapping.

2. Position formation

Do	Review
brainstorm, policy-prfaq, pre-mortem, research-plan	see row 5

Roadmap: options matrix.

3. Evidence base

Do	Review
energy-policy-search, energy-academic-search, energy-data-search, macro-data-search, literature-review, energy-data-analysis, document-distill, empn-paper-distill, think-tank-monitor, think-tank-to-csv, eu-institution-monitor, industry-monitor, media-monitor, academic-monitor, linkedin-monitor, x-monitor, tooling-monitor, energy-modelling-monitor	see row 5

Review gaps: data-replication-check. Roadmap: ECJ / case-law search · regulator decisions · impact-assessment review · scenario comparator.

4. Public-facing output

Do	Review
policy-brief, presentation, html-deck, deck-animated-stage, empn-carousel, html-visual, create-chart, cover-image, empn-layout, publishing-website, tracker-init, tracker-scaffold, tracker-design, tracker-update-data, tracker-publish	editorial-review-structure, editorial-review-prose, policy-brief-validate

Review gaps: tracker-design-review · embed-iframe-check. Roadmap: op-ed · talking points / meeting briefing · chart / visual review · translation EN↔FR/DE.

5. Quality gates (cross-phase)

Do	Review
—	fact-check, adversarial-review, advanced-elicitation

6. Support (transversal)

Do	Review
help, tutorial, project-init, project-status, folder-audit, content-audit, project-audit, contact-profile, contacts-enrich, contacts-commit, contacts-audit, correct-course, project-retrospective, skill-builder, skill-sender, skill-feedback, skill-idea, live-artifact, email-triage	n/a

Roadmap: meeting-note template · glossary maintainer.

Triggering a skill

You don't call skills by name. You describe what you want and Cowork routes you to the right skill based on trigger phrases declared in each SKILL.md. Examples:

• "draft a brief on capacity mechanisms" → policy-brief
• "what's new on the European Grids Package" → eu-policy-research
• "find me the academic literature on carbon pricing pass-through" → energy-academic-search
• "who are the rapporteurs on EPBD recast" → eu-policy-research

If you're not sure which skill applies, type help in any conversation — it routes you live.

Per-skill detail

Per-skill trigger phrases, full descriptions, and hand-offs live in each SKILL.md under plugins/empn-core/skills/. Or type help in Cowork — it enumerates the catalogue live and routes to the right one.

Validation and benchmarks

Each SKILL.md carries three mechanically-maintained frontmatter fields:

• last_audited — when the skill was last consistency-checked (paths, hand-offs, shared-doc references).
• last_benched — when the skill was last run against its benchmark prompts and scored.
• bench_status — pass, fail, or untested.

A scheduled remote agent runs repo-audit and skill-benchmark on a cadence; both fields are backfilled on next run, so a missing value reads as never / untested, not a failure.

Aggregate counts are auto-generated in the repo README.md between the SCOREBOARD-START / SCOREBOARD-END markers. The current snapshot:

Last bench run: none yet — scoreboard populates on the next skill-benchmark run. Status: 0 pass · 0 fail · 70 untested (out of 70 skills). Coverage: at the current --random cadence, every skill is benched every ~9 days; --changed runs cover any skill modified since the last run.

When the benchmark surface stabilises, a live status table will replace the static snapshot on this page.

Trust framing

Skills are not magic. They encode methodology — what sources to consult, what folder to write to, what voice to use — and they keep the model honest by linking out to shared docs at runtime. They do not replace judgement. Treat any skill output as a draft you review; the Quality gates row (fact-check, adversarial-review) exists for exactly that purpose.

The benchmark catches obvious regressions; it does not catch subtle factual drift. Always read what a skill produces before you ship it.