bvandeusen
5fbee18a94
Plan #825 (T2 — Issues task_kind) shipped S1–S4 but its S5 docs slice never landed, so every behavioral surface the plugin pushes to the agent still described the pre-kind convention ("tag `issue`" on a create_note). Result: agents fixed bugs without reaching for kind=issue and dumped the work as logs on unrelated open tasks. - _INSTRUCTIONS: rewrite the "record a problem" bullet to create_task(kind="issue") with symptom→cause→fix + arose_from_id / system_ids, and an explicit "not a work-log on an unrelated task"; add Issue + System to the hierarchy section. - skills/systematic-debugging, verification: drop "tag `issue`" / create_note-issue, point at create_task(kind="issue"). - skills/using-scribe: add issues/systems to the entity list + reflex #6. - hooks/scribe_static_context: fix → its own issue on the keyless floor. Instance-agnostic, prose-only; no schema or tool-behavior change. Pairs with always-on rule #118. Issue: #855. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2.0 KiB
name, description
| name | description |
|---|---|
| systematic-debugging | Use when diagnosing a bug, failure, or unexpected behavior — investigate methodically instead of guessing. Triggers on "why is this failing/breaking", a stack trace, a flaky test, or any "it should work but doesn't". On resolution, capture the issue in Scribe so it isn't re-debugged from scratch. |
Systematic debugging
Find the root cause, don't patch the symptom. Move one step at a time — a guessed fix that "seems to work" often just moves the bug somewhere else.
Recall first
Before digging in, search Scribe for the symptom — a prior issue
(list_tasks(kind="issue") or search) may already hold the cause and the fix.
Don't re-debug what's already solved.
The loop
- Reproduce — get a reliable, minimal repro. If you can't reproduce it, you can't confirm you fixed it.
- Observe — read the actual error / log / state. Don't theorize past the data you have.
- Isolate — narrow to the smallest failing case; change one variable at a time so each result actually means something.
- Hypothesize → test — state the single most likely cause, then test that one thing. Confirm or rule it out before moving on; don't stack guesses.
- Root cause — keep going until you can explain why it failed, not just what made it stop. "It works now" without "because X" is unfinished.
- Fix + verify — fix the cause, then re-run the repro to confirm it's gone.
Capture the issue (so it's findable)
When resolved, record it in Scribe as its own issue (create_task(kind="issue")):
symptom → root cause → fix → how it was verified in the body, optionally
linked to the task it arose from (arose_from_id) and the subsystem it touches
(system_ids). Even a problem fixed in passing is worth two lines — that's how
the next person (or you) avoids re-deriving it. Record it discretely; don't bury
it as a work-log line on an unrelated open task. If the work was already tracked
as its own task, log the resolution there and set it done.