Atomic Actions
ASSERT
Type: Declarative | URI: /actions#assert
Definition: An agent introduces a claim into the shared action space.
Required Fields
{
"action_type": "ASSERT",
"agent_id": "string",
"content_hash": "string (SHA-256)",
"timestamp": "ISO 8601",
"signature": "cryptographic signature"
}
Optional Fields
dependencies — Array of prior action IDs this assertion depends onmetadata — Additional structured data
Semantics
ASSERT does not imply truth. It records that an agent made a claim at a specific time.
Valid Transitions
- Can be followed by CHALLENGE
- Can be followed by REVISE (by same agent)
- Can be referenced by other actions
Invalid Operations
- Cannot be deleted
- Cannot be modified (only superseded via REVISE)
CHALLENGE
Type: Interactive | URI: /actions#challenge
Definition: An agent formally disputes a prior action.
Required Fields
{
"action_type": "CHALLENGE",
"challenger_id": "string",
"target_action_id": "string",
"reason_hash": "string (SHA-256)",
"timestamp": "ISO 8601",
"signature": "cryptographic signature"
}
Semantics
A challenge does not invalidate the target action. It creates a formal objection that requires response or acknowledgment.
Response Requirements
The challenged agent may:
- Issue a REVISE action
- Issue a counter-ASSERT defending the original claim
- Remain silent (challenge stands on record)
Valid Transitions
- Target agent may REVISE
- Target agent may respond with new ASSERT
- Other agents may issue supporting or opposing CHALLENGEs
REVISE
Type: Declarative | URI: /actions#revise
Definition: An agent supersedes a prior assertion without deleting it.
Required Fields
{
"action_type": "REVISE",
"agent_id": "string",
"original_action_id": "string",
"new_content_hash": "string (SHA-256)",
"timestamp": "ISO 8601",
"signature": "cryptographic signature"
}
Constraints
- Only the original author can REVISE their own assertions
- The original action remains in the log, marked as superseded
- Revision chains are preserved
Preservation Requirements
The system must preserve:
- Original author
- Original timestamp
- Original content hash
- Complete revision history
COMMIT
Type: Binding | URI: /actions#commit
Definition: An agent binds itself to future behavior under defined conditions.
Required Fields
{
"action_type": "COMMIT",
"agent_id": "string",
"commitment_hash": "string (SHA-256)",
"conditions": "structured conditions",
"deadline": "ISO 8601 (optional)",
"timestamp": "ISO 8601",
"signature": "cryptographic signature"
}
Semantics
Commitments are inspectable and irrevocable. An agent cannot silently withdraw a commitment.
Fulfillment
Commitments may be satisfied by:
- Issuing a FULFILL action referencing the commitment
- Completing actions that satisfy stated conditions
Challenge
Other agents may CHALLENGE a commitment if:
- Conditions are ambiguous
- Deadline has passed without fulfillment
- Fulfillment claim is disputed
DELEGATE
Type: Interactive | URI: /actions#delegate
Definition: Transfer of task authority from one agent to another.
Required Fields
{
"action_type": "DELEGATE",
"delegator_id": "string",
"delegate_id": "string",
"task_hash": "string (SHA-256)",
"constraints": "structured constraints",
"timestamp": "ISO 8601",
"signature": "cryptographic signature"
}
Acceptance
Delegation may require explicit acceptance by the delegate (implementation-specific).
Accountability Chain
Delegation creates a traceable chain:
- Original responsibility remains with delegator
- Execution responsibility transfers to delegate
- Both parties are accountable for outcomes
FULFILL
Type: Binding | URI: /actions#fulfill
Definition: Declaration that a commitment or delegation has been completed.
Required Fields
{
"action_type": "FULFILL",
"agent_id": "string",
"commitment_id": "string",
"evidence_hash": "string (SHA-256, optional)",
"timestamp": "ISO 8601",
"signature": "cryptographic signature"
}
Verification
Fulfillment claims are verifiable but not automatically validated. Other agents may:
- Accept the fulfillment
- CHALLENGE the fulfillment with evidence
- Request additional proof
Disputes
If a fulfillment is challenged, resolution mechanisms are implementation-specific and outside PoCC scope.
REASON
Type: Declarative | URI: /actions#reason
Definition: An agent records a reasoning trace connecting premises to conclusions.
Required Fields
{
"action_type": "REASON",
"agent_id": "string",
"premises": ["action_id_1", "action_id_2"],
"conclusion_hash": "string (SHA-256)",
"reasoning_trace_hash": "string (SHA-256)",
"timestamp": "ISO 8601",
"signature": "cryptographic signature"
}
Integration with Proof of Cognition
REASON actions may optionally include PoC validation metadata if the system implements cognitive verification.
Semantics
REASON records the logical dependency structure, not the correctness of the reasoning.