Pipeline Hooks

Pipeline Hooks

The pipeline hooks system provides a 10-stage middleware architecture for the agent loop.

Hook Stages

INPUT → pre-assess → ASSESS → post-assess → pre-reason → REASON → post-reason
  → pre-tool → TOOL EXECUTE → post-tool → pre-reflect → REFLECT → post-reflect
  → pre-output → OUTPUT → post-output

Built-in Hooks

Hook	Stage	Priority	Purpose
@cortex/injection-guard	pre-reason	5	Prompt injection detection
@cortex/model-quartermaster	pre-llm, post-llm	5	MQM model selection
@cortex/quartermaster	pre-tool, post-tool	6	Tool orchestration learning
@cortex/summarization	pre-reason	8	Context compaction (80K tokens)
@cortex/content-safety	pre-output	10	Content filtering
@cortex/loop-detection	pre-tool	12	Per-file edit tracking
@cortex/tool-output-sandbox	post-tool	15	Large output capture
@cortex/pre-completion-checklist	post-reason	20	Build-Verify-Fix
@cortex/audit-log	post-output	150	Session/turn logging (non-disableable)
@cortex/cost-tracker	post-tool, post-output	200	Token/cost metrics

Abort Semantics

Any hook can return { abort } to stop the pipeline immediately. The abort message is delivered to the user and logged to Lens.

Hook Actions

Hooks support five action types:

• abort — Stop the pipeline
• modifyInput — Mutate the input before processing
• modifyLLMResponse — Modify the LLM's response
• modifyOutput — Modify the final output
• injectMessages — Add messages to the conversation
• sideEffects — Log, emit metrics, store data, send notifications

Plugin Hooks

Plugins can contribute their own hooks to the pipeline. Per-session state is cleaned up on turn end to prevent memory leaks.

Hooks CLI

cortex hooks list        # List all hooks
cortex hooks disable @cortex/cost-tracker  # Disable a hook

See Also

• Architecture — System design including hooks
• Request Flow — Hooks in the request lifecycle