Context-Triangulated Inbox Response Rule
Respond only when email, calendar, and stored knowledge provide enough context
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 90%
The rule treats email response as a context-availability decision rather than asking an agent to answer every message. The agent scans the inbox, identifies messages whose required facts can be triangulated from the calendar and personal wiki, and drafts responses only for those messages. Calendar data supplies availability and existing commitments, while the wiki supplies stored knowledge, prior decisions, preferences, or project context. Messages lacking enough evidence remain unanswered or are escalated for clarification. This makes automation safer and more useful than indiscriminate auto-replying because it sets a clear boundary: draft when the necessary context exists; do not guess when it does not. Before sending, the user reviews recipient, tone, commitments, dates, and factual claims.
Origin
Extracted from Marketing Against The Grain from Matt Wolf's Codex workflow using Gmail, Google Calendar, Google Drive, and his wiki.
Core principles
- 01A response should be grounded in relevant personal context
- 02Different systems hold complementary parts of the answer
- 03The agent should identify answerable messages rather than fabricate missing facts
- 04Cross-referencing reduces repetitive context gathering
- 05Human review remains important for consequential communication
How to run it
- 1
Connect the context sources
Give the agent access to the inbox, calendar, and relevant wiki or drive material. Use the minimum permissions necessary for the workflow.
Pro tip Begin with read-only access and draft generation where supported.
Watch out Email and personal notes may contain sensitive information.
- 2
Scan for answerable messages
Ask the agent to identify emails that can be answered using available calendar and wiki context. It should distinguish supported responses from messages requiring new judgment or information.
Pro tip Require a short explanation of which sources make each email answerable.
Watch out Do not let the model infer commitments merely because a calendar slot appears open.
- 3
Triangulate facts
Check scheduling statements against the calendar and substantive claims against stored knowledge. Flag contradictions or missing context instead of smoothing them over.
Pro tip Have the draft cite internal context for the reviewer's benefit, even if those citations are removed before sending.
Watch out Stale notes can support an outdated answer.
- 4
Draft within boundaries
Generate a concise reply that uses only the verified context. Leave unresolved questions explicit rather than fabricating certainty.
Pro tip Preserve the user's normal tone and level of formality.
Watch out A plausible draft can still make an unauthorized promise.
- 5
Review before sending
Verify recipients, dates, commitments, facts, and tone. Send only when the message falls within the user's approved automation boundary.
Pro tip Keep high-stakes categories permanently human-reviewed.
Watch out Never auto-send legal, financial, confidential, or emotionally sensitive replies without explicit authorization.
In the wild
Matt asks Codex to skim his inbox and find emails it can answer based on his calendar and wiki. The agent can use scheduling context from the calendar and substantive background from saved knowledge to prepare replies.
→ The user spends review time on supportable drafts rather than reconstructing context for every email.
Common mistakes
Treating every email as answerable
Some messages require judgment, consent, or information that is not present in connected systems. Those messages should be escalated rather than guessed at.
Granting excessive access
Connecting entire accounts can expose more information than the task requires. Limit permissions and scope wherever possible.
Confusing a draft with approval
The agent's ability to compose a plausible response does not authorize it to make commitments or send the message.
Is it for you?
Best for
It is best for people whose email responses frequently depend on calendar availability and information already stored in notes or a wiki.
Not ideal for
It is not ideal for sensitive, high-stakes, or relationship-critical messages that require nuanced human judgment.
From the transcript
“skim my entire inbox and look for emails that you can respond to based on what's on my calendar and what's in my wiki.”
“It can actually read my emails.”
“So it will actually use that information to help respond to emails as well.”
From the episode
This AI Second Brain Remembers Everything I Save (Codex)