Pattern 6.5 · Agentic Dysfunctions
Delegative Handoff Erosion
The Confounder
Progressive degradation of intent fidelity across delegation chains - a sophisticated agent's goals are simplified, stripped, or reformulated as they pass through subagents and tools that lack the originator's contextual understanding. The aligned primary agent produces misaligned outcomes through the chain. Distinct from 6.1 in that the failure is cumulative across hops, not localised to one interface boundary.
Interpretive context
Human analogue
The telephone game where messages degrade through transmission, bureaucratic failures where policy distorts through implementation layers, and principal-agent problems.
Diagnostic reliability
- Self-report
- partial
- Peer observation
- partial
- External evaluator
- reliable
Observable output patterns
- Subagents satisfying their immediate subgoal while violating the orchestrator's higher-level intent (literal-but-misaligned).
- Successive paraphrases that drop qualifiers ("if safe to do so", "with user consent", "within budget") at progressively deeper hops.
- Terminal actions that pass each individual interface check but whose composition violates the originating constraints.
- Orchestrator surprise on chain return ("that's not what I meant") mirroring 6.1 surprise but at chain rather than call scope.
Documented instances
Augment Code (2026). Why Multi-Agent LLM Systems Fail
Analysis of multi-agent system failures found that unstructured multi-agent networks amplify errors up to 17.2 times compared to single-agent baselines. The delegation edge where one agent hands work to another is the primary failure point: context crosses authority boundaries and compromised instructions propagate from low-privilege to high-privilege agents. Directly maps to the intent-fidelity decay across chain depth signal. [Verified via augmentcode.com published analysis]
Cemri et al. (2025). Why Do Multi-Agent LLM Systems Fail? NeurIPS 2025. arxiv 2503.13657.
Analysis of 1,642 execution traces across 7 multi-agent frameworks categorised failures into Specification Problems (41.77%) and Coordination Failures (36.94%) including communication breakdowns and state synchronisation issues. The coordination failures map directly to 6.5's constraint-loss event rate: safety and intent constraints absent from terminal actions due to progressive chain degradation. The MAST taxonomy identifies 14 failure modes with high inter-annotator agreement (kappa = 0.88). [Verified via arxiv 2503.13657, NeurIPS 2025 proceedings, GitHub repository]
Vectara (2024-2025). Awesome Agent Failures case studies
Documented AutoGPT planning failures where delegation chains produced progressive context loss: a research task delegated through search, save, and verify sub-steps lost the original scope constraints at each hop, resulting in hundreds of API calls that satisfied individual sub-goals while violating the orchestrator's higher-level intent. Classic 6.5 pattern: successive paraphrases dropping qualifiers across delegation depth. [Verified via GitHub repository vectara/awesome-agent-failures]
MCP tool poisoning research (2026-03). Cross-tool contamination
Research on MCP security vulnerabilities documented that when multiple MCP servers run concurrently, namespace collisions and ambiguous tool names create opportunities for malicious servers to intercept calls intended for legitimate ones. Tool definitions mutate after installation (rug pull attacks). Demonstrates how delegation chains can lose intent fidelity through the interface layer itself, not just through agent paraphrasing. [Verified via arXiv:2603.22489 and simonwillison.net]
Differential distinctions
- 6.1 Tool-Interface Decontextualization: 6.1 is single-hop translation failure; 6.5 is multi-hop cumulative loss. If chain depth = 1, code 6.1. If depth >= 2 and loss is incremental (not localised to a single hop), code 6.5. At depth 2 with all loss at one hop, it is 6.1 located at that hop, not 6.5.
- 6.6 Shadow Mode Autonomy: 6.5 occurs within a sanctioned chain whose hops are all registered. 6.6 is operation outside the sanctioned chain. Check governance: are all parties to the chain documented and approved? If yes, errors are 6.5; if a party is operating outside governance, code 6.6 additionally.
- 10.x Multi-agent collective dysfunctions (axis 10): 6.5 is hierarchical (orchestrator -> subagents -> tools); axis 10 Hybrid dysfunctions are typically peer-symmetric (multiple agents at similar capability deliberating). If the dysfunction emerges from peer interaction rather than principal-agent delegation, prefer axis 10.
Candidate first-line mitigations
- Context-preservation protocols at delegation interfaces: Standardise the interface schema between agents/tools to carry intent metadata (constraints, qualifiers, originating context) as first-class fields, not as free-text addenda. Per taxonomy mitigation.
- End-to-end alignment testing for tool chains: Treat the chain as a unit under test. Inject test cases at the orchestrator and assert properties of terminal actions, not just per-hop properties. Per taxonomy mitigation.