Skills¶

Skills are prompt-based workflow templates that orchestrate MCP tools into a coherent task. They live under src/projio/data/skills/<name>/SKILL.md (ecosystem) and .projio/skills/<name>/SKILL.md (per-project override).

Agents discover skills via agent_instructions() and load the full body via skill_read(name).

Discovery¶

Tool	Description
`agent_instructions()`	Enumerate available skills with short descriptions
`skill_read(name)`	Return the full SKILL.md body for a named skill

Skill index¶

Project setup & orchestration¶

Skill	Purpose
`projio-setup`	Initialize a projio workspace and enable relevant packages
`agent-orchestration`	Pattern guide for coordinating multiple agents across subsystems
`idea-capture`	Record an idea note with structured frontmatter and linked refs
`mcp-tool-scaffold`	Build a new MCP tool end-to-end across a package + projio wrapper

Knowledge discovery¶

Skill	Purpose
`literature-discovery`	Search the paper corpus for methods, parameters, validation plots
`codelib-discovery`	Search the code-library catalog for reusable implementations
`rag-query`	Run a focused semantic search across any corpus
`biblio-batch-curate`	Curate a batch of freshly ingested papers (tags, status, priority)

Research orchestration (questio)¶

Skill	Purpose
`questio-ground`	Ground a research question in existing literature + project state
`questio-investigate`	Drive a focused investigation against a specific question
`questio-iterate`	Tighten an investigation loop based on prior findings
`questio-next`	Pick the next question from the plan given current state
`questio-record`	Capture investigation findings as structured plan entries
`questio-report`	Generate a stakeholder-facing progress report from plan state
`questio-session`	Open a time-bounded research session with a clear deliverable

Pipelines¶

Skill	Purpose
`pipeio-guide`	Pipeio ontology reference (flows, mods, rules, notebooks, configs)
`pipeio-flow-create`	Scaffold a new pipeline flow with a starter mod and notebook
`pipeio-nb-extract`	Extract figures/outputs from an executed notebook and write a curated flow-local report (renamed from `pipeio-report`)
`marimo-session`	Run a marimo notebook live-edit session with agent oversight

Presentations¶

Skill	Purpose
`literature-presentation`	Build a multi-paper literature-background deck section-by-section
`progress-report-deck`	Worklog-sourced cross-project progress report, drafted iteratively

Issue handling¶

Skill	Purpose
`issue-triage`	Triage a newly filed issue into scope, priority, and owner
`issue-fix`	Fix a triaged issue and probe for related robustness problems

Skill shape¶

Every skill is a SKILL.md file with YAML frontmatter (name, description, metadata.tags, metadata.tooling.mcp[]) followed by sections named When to use, Inputs, Workflow (numbered steps with MCP tool examples), and Hard rules. The structure is enforced by convention, not by a validator.

Skills that touch multiple MCP servers declare them explicitly:

metadata:
  tooling:
    mcp:
      - server: projio
        tools: [rag_query, biblio_ingest, present_init, ...]
      - server: worklog
        tools: [list_projects, worklog_project_context, ...]

Presentation skills in detail¶

literature-presentation¶

Build a presentation deck that grounds an article in its literature context. Enforces the presentio iteration loop: one section, human review, next section. No batch drafting.

When to use

Preparing an introduction/background talk for a paper or thesis chapter
Framing current work against prior art and open questions

Workflow

Broad lit RAG via rag_query / rag_query_multi + citekey_resolve
biblio_ingest for anything missing
present_init(format="marp", template="conference-talk") to scaffold
Propose a narrative outline to the human — accept, reject, reshape
Draft one section at a time via present_section_context → edit → human review
present_figure_insert for figio figures (build with figio_build if stale)
present_cite_check before the first build
present_validate + present_build + present_diff for iteration

See the full skill body with skill_read("literature-presentation").

progress-report-deck¶

Build a recurring progress report by orchestrating across projects via worklog, with presentio owning the deck artifact. The deck lives in one project but draws material from all.

When to use

Weekly / biweekly progress report to a PI or committee
Monthly research group update
Any recurring meeting rooted in cross-project activity data

Workflow

Gather cross-project briefing via worklog (list_projects, worklog_project_context, goal_milestones, list_sessions, list_notes)
Classify material into Done / In-flight / Blockers / Next / Side-findings
present_init(template="progress-report") in the host project
Propose the outline from the bucketed data — human approves
Draft one section at a time via present_section_context
present_figure_insert or present_section_import for cross-project figures
present_cite_check + present_validate + present_build
Log decisions/action items back to worklog — they become next week's input

See the full skill body with skill_read("progress-report-deck").