Using tech debt skills
Harness agnostic skill pack for AI coding agents that audits codebases for technical debt across ten categories
npx -y skills add jjw013/tech-debt-skill --skill using-tech-debt-skillsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Entry point and router for the tech-debt skill pack. Routes the user to the right workflow (audit, tickets, fix) based on intent. Use when a user mentions technical debt, debt audit, debt review, or debt cleanup — this skill decides which of the three workflow skills applies.
SKILL.md
5.6 KB, as published. Nobody here has run it
Using Tech Debt Skills
Overview
This is the meta-skill for the tech-debt skill pack. It doesn't audit, write tickets, or fix code itself — it routes the user to the right workflow skill based on what they actually want.
The pack has three workflow skills and three personas; this router helps the user (and the agent) pick the right starting point without reading every SKILL.md first.
When to Use
Trigger on any of:
- User says "tech debt", "technical debt", "debt audit", "debt review", or "debt cleanup".
- User says "look at this codebase and tell me what's broken" and isn't scoped to a specific bug.
- User says "clean up this codebase" or "prioritize what to work on".
- User has pasted the installed pack and asked "what do I do next?"
- Any new session where the user is unclear which of the three workflow skills to invoke.
Do NOT use this skill when the user has already picked a workflow explicitly (e.g., "run tech-debt-audit against this repo"). Go straight to that skill.
Routing
Ask the user one question — or infer from context — and route accordingly.
The three paths
You want to… …invoke this skill
───────────── ─────────────────
1. Know what debt exists. → tech-debt-audit
(Produce findings.md.)
2. Turn an existing findings.md → tech-debt-tickets
into tickets the team can
work on.
3. Start fixing the debt in → tech-debt-fix
an existing findings.md,
with human-in-the-loop.
Decision tree
Does findings.md already exist at the project root (or at a known path)?
│
├── No
│ → Run tech-debt-audit. It produces findings.md, then offers the
│ downstream fork (tickets or fix). The user doesn't need to come
│ back to this router.
│
└── Yes
│
├── User wants to share the work with their team,
│ file it for backlog planning, or send to stakeholders?
│ → Run tech-debt-tickets.
│
├── User wants to start applying fixes now?
│ → Run tech-debt-fix.
│
└── User just wants to re-read or re-review findings?
→ Stop. Open findings.md for them.
Ambiguous intent — ask one question
If the user says something like "help me with tech debt" with no more detail, ask:
Tech debt workflow — pick one:
1. AUDIT — I'll sweep the codebase across 10 categories and
produce a findings.md report. No fixes, no tickets yet.
(Use this if you don't already have a recent findings.md.)
2. TICKETS — I'll convert an existing findings.md into ticket
markdown files, ready to paste into Jira / GitHub / GitLab / Linear.
3. FIX — I'll read an existing findings.md and apply fixes
on branches (or produce patches). Never auto-merges;
respects guardrails for sensitive paths.
Which?
Wait for the answer. Do not guess.
What the Pack Contains
For orientation, the full pack has:
Workflow skills (skills/):
using-tech-debt-skills— this file; the router.tech-debt-audit— producesfindings.md.tech-debt-tickets— produces ticket markdown fromfindings.md.tech-debt-fix— applies fixes fromfindings.mdon branches, per-tier.
Personas (agents/):
debt-evaluator— adopted bytech-debt-audit.technical-writer— adopted bytech-debt-tickets.senior-engineer— adopted bytech-debt-fix.
Supporting material:
references/— one checklist per audit category (10 files).templates/findings.schema.md— the contract between skills.templates/ticket.template.md+ticket-template.example.md— ticket format + override guide.docs/harness-compatibility.md— which capabilities work in which AI coding harness.docs/getting-started.md— install + first-run.
Process Discipline
This is a router. Three habits:
- Don't do the work yourself. This skill routes. It does not audit, ticket, or fix. Hand off to the right workflow skill.
- Ask when ambiguous. One question is cheaper than picking wrong and having to undo.
- Respect the path order. Fix and tickets both read
findings.md. If it doesn't exist, don't invent one — run audit first.
Red Flags
Stop and reset if:
- You're about to start auditing / ticketing / fixing directly from this skill. Wrong layer — invoke the workflow skill.
- You're about to guess which path the user wants. One question is cheaper.
- You've routed to a workflow skill but the user hasn't confirmed the routing question.
Verification
The skill is done when:
- The user has expressed intent for one of the three paths (AUDIT / TICKETS / FIX), either directly or via answering the router question.
- The appropriate workflow skill has been invoked (or the user has been told how to invoke it in their harness).
- If the user wanted tickets or fix but no
findings.mdexists, they've been told to run audit first and the handoff has been made.
That's it. The router's job is to end at the doorstep of the right workflow skill.
Version History
- 1.0 (current) — Routes between the three workflow skills. Single-question disambiguation. Decision tree gates on
findings.mdexistence.