{"slug":"systematic-debugging-10","title":"systematic-debugging","summary":"Use when encountering a bug, test failure, or unexpected behavior, before proposing fixes","platform":"Claude","tags":[],"authorName":"LLM Mart","authorSlug":"llm-mart","score":0,"source":"github","price":null,"verified":false,"createdAt":"2026-09-15T18:23:36.039953Z","repo":{"url":"https://github.com/GanyuanRan/Aegis","stars":1224,"forks":53,"license":"MIT","updatedAt":"2026-09-20T11:25:37Z"},"bodyHtml":"
\n

name: systematic-debugging\ndescription: \"Use when encountering a bug, test failure, or unexpected behavior, before proposing fixes\"

\n

Execute

\n

Bug, failure, or unexpected behavior:

\n
    \n
  1. Isolate — read error, reproduce, inspect the diff, and drill upward through diagnostic layers:\nL1 symptom → L2 logic → L3 system → L4 architecture → L5 cross-system\ncontract → L6 platform → L7 spec gap. Layers are observation altitudes,\nnot one causal chain; the causal shape at the stop altitude is classified\nexplicitly before any root claim. Stop only when causal proof accounts\nfor the recurrence generator or reaches a T-class boundary.
    2. \n
  2. Identify owner — compare working behavior, trace the bad value, locate the\ncanonical owner, and treat duplicate owners as a finding.
    3. \n
  3. Decide before editing — Before fixing, run Patch-Shape Triage and Ripple Signal Triage when shared logic,\ncontracts, fallbacks, adapters, producer/consumer seams, or source-of-truth\nboundaries are involved. Surface Change Necessity for any new source-code\npath or non-trivial source edit. Run Minimality Check for a new branch,\nfallback, adapter, owner, or compatibility path, and Pre-Edit Complexity\nCheck for an overloaded owner or complexity growth. After Change Necessity\nselects code-change and before the first repair edit, own the TDD Route for\nthe repair slice per test-driven-development (off default; strict on\nbehavior/bugfix/shared/contract/persistence/permission/migration risk).
    4. \n
  4. Prove — test one hypothesis with the smallest reproduction or\nverification. A failing test first is required only by a recorded\nTDD Route: strict; with TDD Mode: off, do not require a failing test or\nRED/GREEN cycle. Three failed fixes means stop and question architecture.
    5. \n
  5. Repair and close — fix minimally at the canonical owner, verify in\nproportion to risk, review architecture, and close both repair and\nretirement tracks. If any symptom remains, stop and diagnose it separately.
    6. \n
\n

Done: confidence ≥ B, causal status matches recurrence evidence or an external\nterminal, tracks explicit, no H signal, and required D evidence passes.

\n

Core invariant

\n

Find root cause and fix the bug class at its canonical owner. A minimal fix is\nnot the smallest textual diff; it is the smallest sufficient owner-level repair.

\n

Quick bug lane

\n

For a low-risk, reproducible, single-owner bug with no patch-shape signal, keep\nthe readback compact: Symptom, Reproduction, Root Cause, Change Necessity, Fix Boundary, and Verification. Skip the causal card only when\nthe causal-proof owner's Quick Exit Proof passes.\nQuick bug lane must surface Change Necessity before source edits. One sentence\nmay cover the user-visible need, no-change/non-code option, why code must\nchange, minimum boundary, and an explicit decision token such as\nDecision: code-change. If shared logic, a contract, fallback, duplicate\nowner, consumer patch, or cross-module behavior appears, leave this lane.

\n

Aegis Visibility names the evidence/owner/patch-shape/verification effect.\nPass root cause, avoided misfix, boundary, evidence, complexity, and risk to\nverification-before-completion; no separate receipt.

\n

Diagnose before repair

\n
    \n
  1. Read the complete error/stack and record inputs, environment, versions, and\nsuccess criteria.
    2. \n
  2. Reproduce consistently. If unstable, read\nfeedback-loop-construction.md only when evidence shows intermittent or\ntiming-dependent reproduction and build a bounded loop. Shrink the repro\nto load-bearing elements as the test input, never the fix scope: still drill\nupward; test at the correct seam.
    3. \n
  3. Inspect recent changes and compare a working example. Code is evidence; if\nauthority, glossary, code, and tests disagree, compose\nestablishing-project-context rather than silently redefining a term.
    4. \n
  4. Instrument component boundaries, then trace the bad value toward its source.\nRead root-cause-tracing.md only when the observed bad value is several\ncalls or components downstream from its origin.
    5. \n
  5. State one hypothesis and falsify it with one-variable evidence. Do not stack\nspeculative fixes. End each loop with Goal | DeeperCause | Evidence | Risk/Unknown | Decision.
    6. \n
\n

Canonical-owner and patch-shape gate

\n

Before editing, continue upward unless evidence proves the local site is the\ncanonical owner when the candidate is any of these signals:

\n\n
PatchShape:\nCanonicalOwner:\nUpwardDrillSignal:\nDecision: fix owner | continue investigation | escalate\n
\n

A locally green test does not erase triage; a renamed carrier is not a new direction.

\n

When a repair may reinterpret or retire existing semantics, responsibility,\ncontract, or relationship, name the behavior to preserve, highest-risk\ncounterexample, and material unknown. For each known explicit anchor or\nupstream/downstream reference, state its role and disposition: preserve, rebind\nto the canonical owner, retire with reason, or reject because of conflict.\nLeave unresolved relationships unknown; do not re-infer them downstream.\nBind role before value and retire invalid responsibility, not evidenced carrier\ncapability. This bounded reminder is not a behavior matrix, relationship graph,\nreferential-integrity proof, or exhaustive discovery claim. It adds no artifact,\nTDD risk signal, or regression scope; the existing TDD route owner and\nconfigured/default mode still apply.

\n

If the diagnosis crosses L3, a patch-shape signal fires, a user disputes the\nroot claim, a prior fix leaves a symptom, compound/root topology is plausible,\ntwo or more anchored manifestations of one incident exist, reproduction\nconditions diverge across occurrences, or an\nupstream producer/config/default/contract/spec remains unexcluded, read\nroot-cause-claim-contract.md before claiming a root cause. It is the sole\nowner of the Pre-Claim Gate, causal-closure/falsifier proof, layer-ceiling\nproof, and Causal Topology Gate.

\n

Change Necessity

\n

This decision is behavior-triggered, not prompt-triggered. It applies to any new source-code path. Before that path or a non-trivial source edit, expose the Change Necessity decision (no-change | docs/config-only | code-change | needs-clarification); field detail lives in advanced-debugging-governance.md.

\n

Minimality and owner fit

\n

For any proposed branch, fallback, adapter, compatibility path, or new owner, run Minimality Check (fields in advanced-debugging-governance.md) with verdict sufficient repair | local patch | needs first-principles review, and retire invalid responsibility: a local patch needs a retention reason and retirement trigger. For a new non-ordinary repair surface, run the Existence Check in docs/current/AEGIS_MINIMALITY_REFERENCE.md. If retirement involves old code,\nexternal compatibility, or persistent-state risk, compose\nanti-entropy-governance; it chooses the retirement path but never grants\ndestructive authority.

\n

Before editing an overloaded or mixed-purpose owner, complete Pre-Edit Complexity Check and Pre-Edit Owner-Fit Decision (templates in advanced-debugging-governance.md).

\n

Use using-aegis/references/complexity-governance.md for pressure signals.\nDo not add new-responsibility in place by default. If the safer boundary\nchanges the approved shape, update the plan/spec first.

\n

Repair and proportional verification

\n

Implement one owner fix; no bundled “while here” work. Under strict TDD, create\nthe smallest failing test first. With TDD off, a reproduction is diagnostic\nevidence, not a RED gate or a prerequisite for production edits.

\n

Verification must match the risk:

\n\n

Read advanced-debugging-governance.md before another fix for failed/\npersistent / divergent repair or three failures; for unclear/disputed stop /\nLayer Stop Card / intervention; or plausible compound root. Closeout triggers:\nrepair-added patch-shape; multi-site/one-regression;\nremaining pattern/anomaly/duplicate/wrong-owner/downstream repair;\nuninspected same-symptom fix; open recurrence/unsupported root status;\nmissing compound topology-specific member/anti-disguise proof;\noutside-repo authority; unmigrated\npublished-contract break; undefined spec; missing permission/info. They route H/T/D;\ndetail is not causal proof.

\n

For non-trivial debugging with configured workspace support:

\n
python <aegis-workspace-helper> init --root <target-project-root>\npython <aegis-workspace-helper> new-work --root <target-project-root> ...\npython <aegis-workspace-helper> add-evidence --root <target-project-root> --work <YYYY-MM-DD-slug> ...\npython <aegis-workspace-helper> check --root <target-project-root>\n
\n

Failed attempts use <aegis-workspace-helper> add-attempt; add-evidence is terminal-only.

\n

Fast bug fix or quick bug fix pressure does not skip this: if Ripple Signal\nTriage fires, record it before editing and verify the canonical owner plus\naffected downstream path. Records are advisory, not completion authority.

\n

Closure

\n

Always report:

\n\n

Confirm the reproduction, same-pattern handling, authority, complexity, and\nretirement. Prefix debug logs (e.g. [DEBUG-a4f2]); confirm one-grep removal before close. Confidence: A = direct regression evidence; B = strong evidence\nwith bounded unknowns; C = partial and not resolved.

\n

Trace Digest may summarize audit evidence; never expose chain-of-thought or\nreplace root-cause, rule-effect, and verification evidence.

\n","files":[{"path":"advanced-debugging-governance.md","sizeBytes":8553,"isText":true},{"path":"condition-based-waiting-example.ts","sizeBytes":5054,"isText":true},{"path":"condition-based-waiting.md","sizeBytes":3516,"isText":true},{"path":"defense-in-depth.md","sizeBytes":3650,"isText":true},{"path":"feedback-loop-construction.md","sizeBytes":2479,"isText":true},{"path":"find-polluter.sh","sizeBytes":1528,"isText":true},{"path":"root-cause-claim-contract.md","sizeBytes":13870,"isText":true},{"path":"root-cause-tracing.md","sizeBytes":5615,"isText":true},{"path":"SKILL.md","sizeBytes":10874,"isText":true}],"reviewScore":null,"reviewSummary":null,"trust":{"provenance":"trusted-source-unreviewed","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow.","bodySource":null},"bodyLocked":false,"purchaseUrl":null,"sourceUrl":null,"report":{"provenance":"trusted-source-unreviewed","screen":{"ran":true,"outcome":"clean","suspicious":0,"notes":0,"hiddenCharacters":false},"virusScan":{"engine":"clamav","status":"clean","scannedAt":"2026-09-21T13:50:27.208531Z","sha256":"6071B3DEFEB5B21A9DAB7E6BA9CEA39C32A47D10A18F44550F80A65DCBBD969C","sizeBytes":23915},"review":null,"source":{"repositoryUrl":"https://github.com/GanyuanRan/Aegis","path":"skills/systematic-debugging","license":"MIT","commit":"60321eda49c3b473a183ca0edef5fbfba91eaad1","subtreeSha":"5DD385DC4B1237731D6300573F1007B9AE998B114F2B1227A900E401A719015C","lastSyncedAt":"2026-09-21T13:50:15.063596Z"},"reviewedAt":"2026-09-21T13:50:30.474245Z","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow."},"install":[{"target":"skills-cli","command":"npx skills add https://github.com/GanyuanRan/Aegis/tree/main/skills/systematic-debugging"},{"target":"claude-code","command":"claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install ganyuanran-aegis@llmmart"},{"target":"git","command":"git clone https://github.com/GanyuanRan/Aegis.git"}]}