Skip to content

questio phase 3: create questio skills (next, ground, record, report, session)

Goal

Create the prompt-based skills that compose questio MCP tools with existing subsystem tools to enable full agentic research workflows.

Context

The questio spec (section 6.2 of docs/specs/research-orchestration/design.md) defines 8 skills. This task creates the 5 core operational skills. The questio-validate, questio-ready, and questio-docs-refresh skills are simpler and can be added later.

Skills are markdown files in .projio/skills/ that contain structured prompts. They are discovered by agent_instructions() and executed via skill_read(). See the projio skill system documentation and any existing skills for format reference.

Prerequisite: Phase 2 (questio_status and questio_gap tools exist).

Prompt

Create questio skills in the projio project's .projio/skills/ directory.

Step 1: Read the questio spec sections 6.2 and 8 (operational workflows) at /storage2/arash/projects/projio/docs/specs/research-orchestration/design.md.

Step 2: Check how existing skills are structured. Look at: - Any .projio/skills/*.md files in projio or pixecog - The archived skills at /storage2/arash/projects/pixecog/.projio/skills_archive/ for format examples - The skill_read MCP tool implementation to understand what format skills should be in

Step 3: Create the following skills in /storage2/arash/projects/projio/.projio/skills/:

questio-next.md

Skill prompt that: 1. Calls questio_status() to get current research state 2. Calls questio_gap() for each in_progress or not_started question 3. Cross-references with pipeio_flow_status() to check pipeline readiness 4. Reasons about highest-impact unblocked work considering: - Which milestone unblocks the most downstream milestones/questions - Which pipelines are already close to completion - Which questions have the most evidence already (closer to sufficiency) 5. Outputs a ranked recommendation with rationale

questio-ground.md

Skill prompt that, given a milestone or question ID: 1. Extracts relevant citekeys from questions.yml citations field 2. Calls paper_context() for key papers to get expected methods and values 3. Calls codio_discover() with keywords from the milestone description to find existing implementations 4. Calls rag_query() to search project notes for prior attempts or decisions 5. Synthesizes: recommended approach, expected results, quality criteria, potential pitfalls 6. Outputs a grounding brief the agent uses before starting implementation

questio-record.md

Skill prompt that guides the agent through result capture: 1. Asks which question(s) and milestone the result relates to 2. Validates that the question IDs and milestone ID exist in questions.yml/milestones.yml 3. Creates a result note via note_create(type="result") with proper frontmatter 4. Updates plan/milestones.yml to add the note ID to the milestone's evidence list 5. If the milestone now has sufficient evidence, prompts about updating its status 6. Calls questio_docs_collect() to regenerate plan docs

questio-report.md

Skill prompt that generates a supervisor-ready progress summary: 1. Calls questio_status() for overall state 2. Searches recent result notes (last 7 days by default, configurable) via note_search 3. Identifies milestones completed recently 4. Lists current blockers and next planned actions 5. Formats output as a concise markdown report with sections: - Progress (milestones hit, % change) - Key Results (recent evidence with metrics) - Blockers - Next Steps - Timeline impact (if any milestones are behind)

questio-session.md

Skill prompt that structures a full research session: 1. Orient: call questio_status(), summarize state 2. Plan: call questio-next skill logic to pick work 3. Ground: call questio-ground skill logic for the chosen milestone 4. Present the plan to the user for approval 5. After work is done, prompt to record results via questio-record 6. Close: call questio-report to summarize session accomplishments

Step 4: Ensure each skill file has proper frontmatter: 

---
name: questio-next
description: "Recommend highest-impact research work based on questio status and evidence gaps"
tools: [questio_status, questio_gap, pipeio_flow_status]
---

Step 5: Verify skills are discoverable: check that skill_read("questio-next") returns the skill content.

Step 6: Commit: "Add questio skills: next, ground, record, report, session"

Acceptance Criteria

  • [ ] 5 skill files created in .projio/skills/
  • [ ] Each skill has proper frontmatter (name, description, tools)
  • [ ] Skills reference correct MCP tool names
  • [ ] questio-next includes dependency-aware prioritization logic
  • [ ] questio-ground composes biblio + codio + rag for pre-work context
  • [ ] questio-record includes milestone YAML update step
  • [ ] questio-report produces supervisor-readable output format
  • [ ] questio-session chains the other skills into a full workflow
  • [ ] Skills discoverable via skill_read
  • [ ] Committed

Result

(Filled in after execution)