# Agent Error Attribution > Activate whenever a multi-agent system or single complex agent fails and the question is WHERE and WHY — post-incident analysis of agent failures, debugging multi-step agent pipelines, "the agent did something wrong but I don't know which part broke," designing error taxonomies or failure telemetry for agent systems, or distributing responsibility for a failure across a delegation chain. Fire because the default failure analysis in agent systems is "the model hallucinated" — a non-diagnosis that blames the last visible step and fixes nothing, guaranteeing the failure returns wearing a different output. - Skill: `forexgod21/agent-error-attribution` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add forexgod21/agent-error-attribution` - Raw SKILL.md: https://api.skillmd.com/api/skills/forexgod21/agent-error-attribution/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - License: CC BY 4.0 - Author: Forexgod21 (https://skillmd.com/u/forexgod21) - Updated: 2026-09-21 - Page: https://skillmd.com/skills/forexgod21/agent-error-attribution --- # Agent Error Attribution **Attribution:** YourVisionYourCreation LLC — yourvisionyourcreation.com **Research anchor (verified):** Tomašev, Franklin & Osindero, *Intelligent AI Delegation* (Google DeepMind, arXiv:2602.11865, 2026), whose safe-delegation dimensions include accountability and verifiable task completion. This skill is YVYC-original doctrine: a five-stage failure taxonomy and attribution walk that operationalizes accountability across agent stages and delegation chains. **Doctrine class:** Tier 5 — Frontier (YVYC original, anchored in verified research) --- ## Universal So-What "The model hallucinated" is agent failure analysis at the level of "the patient got sick." Agent failures have anatomy: a wrong memory retrieved, a plan that never matched the goal, a tool called with mangled parameters, a reflection step that approved garbage. Each failure class lives at a different stage, is owned by a different component, and demands a different fix. Attribution is the discipline of finding the ACTUAL failing stage — because a fix applied to the wrong stage is a cost paid for a failure kept. --- ## Core Doctrine ### 1. The Five-Stage Failure Taxonomy Every agent failure is classified to the stage where it ORIGINATED — not the stage where it became visible: | Stage | Failure Class | Signature | |---|---|---| | **Memory** | Wrong, stale, or missing context retrieved; poisoned or corrupted state | The inputs to reasoning were bad before reasoning began | | **Reflection** | Self-assessment approved bad work or rejected good work | The checking layer failed, not the doing layer | | **Planning** | Decomposition wrong, steps misordered, goal misread, constraints dropped | Every step executed "correctly" toward the wrong plan | | **Action** | Tool misuse, malformed parameters, wrong tool selected, output misparsed | The plan was sound; the hands fumbled | | **System** | Infrastructure, timeout, rate limit, permission denial, orchestration bug | No cognitive stage failed; the floor collapsed | The cardinal analytical rule: failures CASCADE forward. A memory-stage failure produces planning-stage symptoms and action-stage wreckage. Attribution walks BACKWARD from the visible failure to the first stage where things were already wrong — the first-error discipline of debug-evidence-protocol, applied to cognition. ### 2. The Attribution Walk For every failure, in order: 1. **Capture the wreckage:** the visible failure, exactly — output, action, or omission 2. **Reconstruct the trajectory:** the full path — what was retrieved, what was planned, what was executed, what was checked — from logs and traces, not from the agent's own summary of itself 3. **Walk backward:** at each stage, was this stage's OUTPUT correct given its INPUT? The first stage whose output was wrong given correct input is the origin. 4. **Classify:** assign the taxonomy class at the origin stage 5. **Verify:** the classification must predict — fixing this stage should prevent this failure; if a replay with the stage corrected still fails, the walk missed an earlier origin ### 3. Attribution Across Delegation Chains When the failing system is multiple agents: - The stage taxonomy applies WITHIN each agent; chain attribution decides BETWEEN agents — which link's defect originated the failure - Three chain-level verdicts, each landing differently: bad SPECIFICATION (the delegator asked for the wrong or impossible thing), bad EXECUTION (the delegatee failed a sound task), bad DECISION (the delegation itself should never have happened — wrong delegatee, wrong time, failed gate) - Responsibility propagates to the origin: a sub-agent executing an impossible specification perfectly is not the failure; the specification is (extending accountability-chain, Tier 3) ### 4. Blameless Toward Agents, Accountable Toward Design - The attribution's product is a DESIGN verdict, not a culprit: the failing stage indicates which component, prompt, tool schema, memory system, or gate needs the fix - "The model is bad" is banned as a terminal diagnosis — if the model underperforms at a stage, the finding is what scaffolding, verification, or task-shaping that stage needs, or evidence the task exceeds current capability and needs human retention - Failure data aggregates without shame: an error ledger by stage and class, because the PATTERN across twenty failures is worth more than the autopsy of one ### 5. The Telemetry Prerequisite Attribution is only as good as the trajectory record: - Each stage logs its inputs and outputs at a fidelity that permits the backward walk — an agent system that logs only final outputs has pre-decided that every failure will be attributed to nothing - The agent's self-narrative ("I checked the data and it looked fine") is testimony, not telemetry — useful as a pointer, never as evidence - Telemetry gaps found during an attribution walk are themselves findings: the next incident should never hit the same blind spot ### 6. The Fix Lands at the Stage - Memory-stage origins get memory fixes: retrieval quality, staleness checks, poisoning defense — not prompt tweaks - Planning origins get planning fixes: decomposition scaffolds, constraint pinning, goal restatement — not tool schema changes - Reflection origins get reflection fixes: independent verification, adversarial checks — not more of the self-review that already failed - The anti-pattern with a name: PATCHING THE SYMPTOM STAGE — adding action-stage guardrails for a planning-stage disease produces agents that execute wrong plans more carefully --- ## Common Failure Modes | Failure | Cause | Correction | |---|---|---| | "The model hallucinated" closes every incident | Non-diagnosis accepted | Five-stage classification, origin required | | Same failure returns in new costume | Symptom stage patched | The fix lands at the origin stage | | Last agent in the chain always blamed | Visible ≠ origin | Backward walk; chain verdicts at the right link | | Attribution impossible after incidents | Output-only logging | Stage-level telemetry as a prerequisite | | Agent's self-report drives the analysis | Testimony treated as telemetry | Logs and traces only; narrative as pointer | | Twenty failures, no pattern seen | Autopsies without a ledger | Error ledger by stage and class, reviewed on cadence | --- ## Non-Negotiables 1. Every failure is attributed to its ORIGIN stage, not its visible stage. 2. "The model hallucinated" is never a terminal diagnosis. 3. Chain failures land on the defective link — specification, execution, or decision. 4. The agent's self-narrative is never evidence. 5. Fixes land at the origin stage or they are symptom patches. 6. Telemetry gaps discovered in one incident are closed before the next. --- *Built by YourVisionYourCreation LLC — yourvisionyourcreation.com* *Research foundation credited above. Licensed under CC BY 4.0*