Capture & contradiction detection
Governed memory would go stale if it only ever held what you wrote by hand. The work itself is where new decisions get made, so Meetless Agent captures the work and turns it into evidence, without letting that evidence pretend to be trusted.
What gets captured
While a session runs, the connector observes and records:
- Prompts you submit.
- Decisions made during the session, including the answers to explicit in-session questions.
- Tool calls the agent makes.
- Documents the agent produces, along with the citations behind them.
All of it lands as evidence, born pending. Capture never promotes anything on its own; see the trust model.
Produced documents get indexed
When a session writes a durable document (a spec, a decision note), Meetless Agent indexes it into the owner’s knowledge base so it becomes retrievable. It is indexed as pending, not canonical: it is searchable immediately, but it does not ground answers until a human accepts it. This is how the corpus grows from real output while staying trustworthy.
Contradiction detection
Capture is not just archival. When a fresh decision contradicts something already in the approved corpus, that is exactly the moment worth catching, and Meetless Agent catches it.
- It surfaces the contradiction rather than silently overwriting either side.
- It can open a human triage case so someone accountable decides which claim wins, on the record.
- It can steer the live session, bringing the conflict to attention while the agent is still working, instead of after the change has shipped.
Why this closes the loop
Injection and Ask put decisions into the agent. Capture and contradiction detection let the agent’s own work feed decisions back, safely. Together they are why governed memory can grow with your project instead of decaying behind it, and why “we quietly changed our mind in a thread” stops silently breaking downstream work.