openclaw - 💡(How to fix) Fix feat(tasks): normalize blocker and failure taxonomy [1 comments, 2 participants]

Summary

Introduce a shared blocker/failure taxonomy for tasks so OpenClaw can distinguish between waiting, blocked, failed, and recoverable transient failure in a machine-readable way.

Problem

Some task states and failure conditions are visible today, but the reason behind a stall or failure is not always normalized.

This limits:

• retry decisions
• audit quality
• user-facing messaging
• future observability

Without a shared taxonomy, different consumers can treat the same condition differently.

Proposal

Define and adopt a normalized set of blocker/failure classes such as:

Blocker-like classes

• approval_required
• external_wait

Failure-like classes

• tool_runtime_error
• provider_error
• transport_error
• config_error
• timeout
• interrupted
• unknown_failure

Exact naming can be adjusted, but the key is to centralize semantics and reuse them across the task system.

Scope

• Add shared blocker/failure classification helpers and types
• Map existing task conditions onto the taxonomy
• Reuse the taxonomy in:
  • audits
  • summaries
  • user-facing status output
• Keep unknown/unclassified paths explicit

Acceptance criteria

• Known blocked/failure conditions map to a shared classification
• Audits and summaries can refer to the normalized class instead of ad hoc prose
• The same underlying condition is not represented differently across multiple consumers
• Future retry logic can branch on classification without bespoke heuristics