paulchinjr.bsky.social / collaborative-problem-solving-skill
an agent skill that brings Dr. Ross Greene's collaborative problem solving framework to engineering tasks for solutions without blame.
Star 0
Fork 0
Watch
4 1 0
.gitignore
README.md
SKILL.md
README.md

Collaborative Problem Solving Skill #

This repository contains a Codex skill for slowing down stalled engineering work long enough to identify the constraint that is actually blocking progress.

The skill is meant for situations where repeated effort is not producing better outcomes: similar fixes keep failing, requirements are unclear, validation is missing, or the next implementation step depends mainly on assumptions.

Purpose #

The skill helps Codex shift from immediate solution generation to evidence-based problem solving.

It encourages a lightweight sequence:

  1. Observe what happened.
  2. Identify what is still unknown.
  3. Classify the limiting constraint.
  4. Choose one measurable next step.
  5. Verify the result.

For more complex or repeated failures, the skill expands into a fuller diagnostic sequence that includes engineering concerns, tradeoffs, and explicit verification.

Engineering Discipline #

This project treats stalled engineering work as a signal that an unmet constraint exists. The constraint may be technical, informational, procedural, environmental, or relational.

The central operating idea is simple:

Before escalating implementation effort, improve the quality of understanding.

In practical terms, that means separating observations from interpretations, avoiding repeated failed actions, making assumptions explicit, and choosing next steps that produce evidence.

Influences And Credit #

This skill is inspired by ideas from Dr. Ross Greene and Janet Lansbury, adapted here into an engineering collaboration context.

Dr. Ross Greene's collaborative problem solving work, represented through Lives in the Balance, emphasizes understanding unsolved problems and unmet constraints before moving to intervention. That framing strongly informs this skill's focus on identifying the constraint before prescribing a fix.

Janet Lansbury's work, including the Unruffled podcast, emphasizes calm observation, respect, and non-blaming interpretation of behavior. That influence appears in the skill's communication guidance: describe observable facts, avoid assigning intent, and treat unexpected outcomes as useful signals rather than personal failures.

Any errors, oversimplifications, or engineering-specific adaptations in this project are the responsibility of this project, not those thinkers.

References:

When To Use #

Use this skill when:

  • A similar failure repeats.
  • The same fix has been attempted more than once.
  • Requirements or success criteria are unclear.
  • Validation is missing, flaky, or inconclusive.
  • Multiple valid options exist and the priority is unclear.
  • The user or agent is frustrated, uncertain, or guessing.

Do not use it for routine edits where the task, success condition, and validation path are already clear.

Testing #

Validate the skill structure:

python3 /home/pchinjr/.codex/skills/.system/skill-creator/scripts/quick_validate.py .

Forward-test the behavior with realistic stalled-work prompts. A good result should:

  • Separate facts from theories.
  • Identify the likely constraint.
  • Ask only high-value questions.
  • Recommend one measurable next step.
  • Include an explicit verification method.
  • Avoid using the full diagnostic template for small problems.

Files #

  • SKILL.md - the actual Codex skill instructions.
  • README.md - human-facing project documentation and attribution.