Tech Debt

A single committed table (docs/TECH_DEBT.md) of code-level cleanup that isn't a risk, a shortcut, or an open decision (docs/RISK_REGISTER.md's job) and isn't a future feature (docs/ROADMAP.md's job) -- just things that should get tidied up when there's spare time. Sibling to the risk-register skill, sharing its scaffolding pattern (dedicated skill, YAML row template, JSON schema, validation script) but not its append-only invariant: a tech-debt row is deleted once fixed, not archived, because there is no decision to keep evidence of.

Rules

1. A row is deleted once fixed, not archived. Unlike risk-register's docs/RISK_REGISTER.md, there is no append-only guard here and no technical backstop preventing deletion -- that is the deliberate design, not a gap. A concurrent-edit merge that drops a still-open row alongside a legitimate deletion is a known, accepted limitation of this model; it is not engineered around, because the whole point of choosing this model over risk-register's is to avoid that machinery for low-stakes items.
2. Enforcement is local pre-commit only, not CI-backed, not GitHub-server-enforced. scripts/validate-tech-debt-schema.sh, wired into pre-commit via lefthook, blocks a locally-committed malformed row. A commit made through the GitHub web UI or API, git commit --no-verify, or from a machine without lefthook installed is not stopped by this.
3. Due diligence runs in both directions. Before filing a row, confirm the item is genuinely just cleanup --not a risk, shortcut, or open decision that belongs in docs/RISK_REGISTER.md instead. Before deleting a row (marking it fixed), confirm the fix didn't reveal a real cost or risk; if it did, promote it to docs/RISK_REGISTER.md first (see Workflow) rather than letting the row vanish with the discovery unrecorded.

Prerequisites

• docs/TECH_DEBT.md exists. If this is the first entry in a project and the file doesn't exist yet, create it with a # Tech Debt title, an intro paragraph drawing the boundary against RISK_REGISTER.md/ROADMAP.md/aislop's baseline, the enforcement-scope caveat from Rule 2, and an empty table using the columns in assets/templates/tech-debt-row-template.yaml.
• assets/templates/tech-debt-row-template.yaml -- the shape of a single row, illustrated
• assets/schemas/tech-debt-row.schema.json -- the same shape, as enforceable constraints
• scripts/validate-tech-debt-schema.sh, wired into pre-commit (lefthook), so a malformed row can never be committed. No append-only companion (Rule 1).
• jq (schema constraints are read at runtime, never hardcoded -- see Scripts) and shellcheck (CI and pre-commit both gate the validator script), both pinned in mise.toml.

When to Use

• A lint, format, or style violation is noticed that an existing gate doesn't fail the build on (a warning, not an error; or a path a gate doesn't cover) -- the exact situation the seed row in docs/TECH_DEBT.md documents.
• A small, low-stakes cleanup item is noticed while working on something unrelated, and isn't worth fixing in the same change.
• A filed item is actually fixed and its row needs deleting.
• Someone asks what code-level cleanup is outstanding.

When NOT to Use

• Anything with a real cost or risk if left unaddressed -- that's docs/RISK_REGISTER.md (risk-register skill), not this doc. If a tech-debt item turns out to carry one, promote it (see Workflow).
• aislop baseline findings (.aislop/baseline.json) -- already a scored, regression-tracked inventory of code-quality findings with its own mechanism. Don't duplicate a finding aislop already tracks.
• Anything confidential -- this file is committed and visible to everyone with repo access.

Area Taxonomy and Row Schema

area is a small, tech-debt-specific set of values (currently Code quality and Testing), not a closed JSON Schema enum and not the .context/ themes vocabulary. RECOMMENDED reading before filing or resolving a row: see Area Taxonomy and Row Schema for the full value list, the revisit trigger for splitting an overloaded area, and the complete column-by-column row schema.

Workflow

Adding an entry

1. Confirm this is genuinely just cleanup, not a risk, shortcut, or open decision (Rule 3). If it carries a real cost or failure mode if left unaddressed, it belongs in docs/RISK_REGISTER.md instead.
2. Read the current last row's # in docs/TECH_DEBT.md; the new row's # is that value + 1 (rows are freely reused after deletion, so don't assume the max # ever seen -- just the max currently present).
3. Write a self-contained Description -- don't assume the reader has whatever prompted this open in front of them. Never cite a .context/ file as where the extra detail lives -- .context/ is gitignored local scratch, so a reference to it may be dead for every reader but the one who filed it. Inline the essential facts directly, or point at a committed doc (an ADR, an MR description, another docs/*.md page) instead.
4. Append the row, matching assets/templates/tech-debt-row-template.yaml's field shapes (Area from the current taxonomy, Effort from its enum, Since as YYYY-MM-DD, Status starting Open).
5. Run scripts/validate-tech-debt-schema.sh before committing (the pre-commit hook runs it too, but check early to avoid a bounced commit).

Resolving an entry

1. Find the row by its # or Item name.
2. Confirm the fix didn't reveal a real cost or risk (Rule 3). If it did, add a row to docs/RISK_REGISTER.md first, with the new row's Description referencing this tech-debt item by name -- since this row is about to be deleted and would otherwise leave no trace of where the risk was first noticed.
3. Delete the row entirely. Unlike risk-register, there is no Resolved state to flip to -- deletion is the resolution.

Relationship to risk-register

The boundary between the two docs is exactly the kind of judgment call people get wrong -- that's the reason this skill's grounding finding (a 2026-07-31 governance-model analysis) exists. If a risk-register row turns out, on inspection, to actually be cleanup rather than a real risk, that's a correction to make via the risk-register skill (amend or, if genuinely mis-filed, raise it there), not a silent move into this doc.

Mindset

• Rule 1 is a deliberate tradeoff, not an oversight. risk-register's append-only guard exists to preserve a decision audit trail; tech debt carries no decision to preserve evidence of. Don't add append-only machinery here "for consistency" -- that would defeat the reason this skill exists as a separate, lighter sibling. See the tech-debt-vs-risk-register governance rationale that grounds this skill for the full reasoning.
• This is a PoC-phase project: proportionate is fine. TYPICALLY a docs/TECH_DEBT.md with one honest row is more useful than an empty one or no doc at all -- RECOMMENDED to ship with whatever real rows exist rather than waiting for a "complete" sweep.
• BY DEFAULT, resolve the risk-vs-cleanup judgment call in Rule 3 by asking "does this carry a cost if nobody ever touches it again?" PREFER filing to docs/RISK_REGISTER.md UNLESS the answer is a clear no.
• Kill criterion: if zero rows beyond the one seeded on launch have been added to docs/TECH_DEBT.md by 2026-10-31 (90 days from creation), delete docs/TECH_DEBT.md and this skill -- the doc isn't earning its keep if nobody is using it. The 90-day window is tied to this skill's 2026-07-31 seeding date, the same audit referenced below.

Anti-Patterns

NEVER add an append-only guard or a Resolved status value to this doc. WHY: Both exist in risk-register to preserve a decision audit trail. Tech debt carries no decision -- adding either here just re-creates the machinery this skill was deliberately built to avoid. BAD: "For consistency with risk-register, let's also block row deletion here." GOOD: Delete the row. That's the resolution mechanism, not a shortcut around one.

NEVER file a real risk or open decision here because it's less ceremony than risk-register. WHY: docs/TECH_DEBT.md has no append-only guard and no audit trail -- a real risk filed here can be silently lost the moment its row is deleted, with nothing to notice. BAD: Filing "Directions API retries could silently exhaust the free quota" as tech debt to skip risk-register's due-diligence step. GOOD: File it in docs/RISK_REGISTER.md, where deletion is blocked and the item can't quietly disappear.

NEVER delete a row without checking whether the fix revealed a real cost or risk (Rule 3). WHY: Deletion is irreversible in practice here -- there's no technical backstop, so a real risk "resolved away" by deletion instead of promoted leaves no trace and no failing check to catch the mistake. BAD: Deleting the lint-violation row the moment the lint is fixed, without checking whether the underlying test gap it was masking is itself a bigger problem. GOOD: Confirm the fix is complete and clean, then delete -- or, if it wasn't, promote to risk-register first.

NEVER cite a .context/ file as where a row's extra detail lives. WHY: .context/ is gitignored local scratch, not a committed, durable record -- a reference to it may point at something no other reader, machine, or fresh clone can ever open, the moment the row is committed. BAD: Description: ... (.context/findings/2026-07-31-foo.md, finding M2) GOOD: Inline the essential fact in the row itself, or reference a committed doc (ADR, MR, another docs/*.md page) if it doesn't fit.

NEVER invent placeholder rows to make the doc look more populated than it is. WHY: A fabricated row is worse than an honest gap -- it looks like real signal to the next reader and wastes their time chasing cleanup that doesn't exist. The 2026-07-31 launch audit found exactly one genuine candidate and shipped with that rather than padding the count. BAD: Adding a vague "clean up some old code somewhere" row just to reach a target count. GOOD: Ship with however many real, verified rows exist -- one is more useful than five fabricated ones.

Scripts

One check -- no append-only companion (Rule 1), unlike risk-register's two:

scripts/validate-tech-debt-schema.sh   # Content: does every row match tech-debt-row.schema.json?

Troubleshooting

Problem                                             | Solution
Pre-commit blocks: a field doesn't match            | validate-tech-debt-schema.sh caught a malformed field --
                                                     | fix it to match tech-debt-row.schema.json's constraints.
Pre-commit blocks: schema required fields mismatch  | The script's column order is out of sync with a schema edit --
                                                     | update validate-tech-debt-schema.sh's expected_fields.
Not sure what the next # should be                  | Read the last row of the table; next # = its # + 1 (# is
                                                     | reused after deletion, so this is the max currently present,
                                                     | not the max ever assigned).
Unsure whether something belongs here or in         | Does it carry a real cost/risk if left unaddressed? -> register
docs/RISK_REGISTER.md                               | row. Is it just cleanup for spare time? -> tech-debt row.

References