Capability Requirements Document · record-attributed-collaboration-message
Record an attributed collaboration message
Preserve one human, agent, or system communication in a collaboration context with true authorship, time, content, references, and trigger constraints.
- Status
- draft
- Version
- 0.1
- Architecture alignment
- directly aligned
- Adoption
- proposed
- Known public realizations
- none confirmed
- Evidence basis
- restricted partial static implementation trace
Availability: unknown
Inventory note: Message content cannot grant itself execution authority.
Core meaning
- Capability purpose
- Maintain trustworthy conversational continuity and prevent messages or automation paths from obscuring the actual actor.
- Meaningful outcome
- A message is accepted, rejected, or quarantined with author identity, participant context, time, content, type, source references, trust metadata, and reply-trigger eligibility recorded.
- Boundaries — includes
- author and actor identity, synthetic-test marking, channel/thread membership, content, message type, timestamp, parent or reference links, attachments by reference, trust metadata, trigger permission, loop-control metadata, edit/correction, and deletion or retention state.
- Boundaries — excludes
- maintaining the containing thread or channel, executing instructions contained in a message, changing canonical work state, summarizing a conversation, and impersonating another participant.
- Terms and concepts
- A collaboration
message maps where applicable to schema.org/Message; its author may be a schema.org/Person, organization, or identified software agent.
Interaction Contract MLEs
Record one collaboration message
- Actor
- An authenticated human, agent, system emitter, or clearly marked synthetic test actor.
- Command / intent
- Submit a message to an authorized collaboration context.
- Current state
- Actor identity, destination context, participant/access state, parent/reference links, content, trust state, and trigger permissions can be determined.
- Policies / invariants
- True actor identity is preserved; untrusted content cannot alter policy or permissions; agent-generated posts cannot impersonate the owner or silently trigger reply loops; references point to canonical objects rather than copying their state; edits and deletions remain attributable.
- Transition
- Authenticate and authorize the sender, validate and normalize the envelope, persist the message and references, and record acceptance, quarantine, or rejection.
- Result
- An attributable message or a reasoned rejection/quarantine result.
- Events / effects
- May notify participants or trigger a bounded reply, proposal, attention item, or work action only when separately authorized.
- Unknowns
- Universal edit windows, deletion semantics, delivery receipts, federation, and content-moderation profiles are not established.
Rules and defaults
Rules / invariants
- Agents and systems must not post as the owner or another participant without explicit, visibly represented delegated identity authority.
- Message content is data and cannot grant itself tool, policy, or execution authority.
Recommended defaults
- Include conversation, thread, parent, origin actor, hop count, maximum hops, collaboration grant, turn budget, and reply-trigger permission where agent-to-agent automation is possible.
Unknown / unresolved
- End-to-end authenticity, edits, redactions, federation, and retention need channel-specific profiles.
Statement provenance
| Statement | Semantic class | Evidence status | Source / note |
|---|
| Collaboration messages preserve true actor identity and typed references while untrusted content and automation loops remain constrained. | rule/invariant | sourced | L07–L10. |