Governed Decisions Log Cloud
The Governed Decisions Log is a versioned, auto-maintained record of governance decisions — surfaced directly as context your coding agent's harness reads, so agents work with an up-to-date picture of what's already been decided instead of repeating a question a human already answered.
What you'll learn
- What is (and isn't) recorded in the log
- How to turn it on for a workspace
- Where the log lives, and how your agent sees it
- The token-cost trade-off, and why it's opt-in
Scope boundary
This is governance-decision records only — adjudications, approved/rejected decisions, resolved incidents, and settings changes. It is not conversational memory, session history, or general-purpose context management. Intutic does not manage your agent's full conversational context; that's a different kind of product, and one this feature deliberately does not attempt.
What's recorded
Each entry in the log is a one-line summary of one governance event:
| Source | What shows up |
|---|---|
| Decisions | Approved/rejected Review Queue decisions |
| Incidents | Resolved governance incidents — category and severity, not the full incident narrative |
| Settings changes | Which workspace settings changed, and by whom (key names only, never values) |
| Detector-finding adjudications | A detector finding ruled TRUE_POSITIVE or FALSE_POSITIVE — the pattern name, never any quoted model output |
Summaries are rendered server-side, deliberately kept to structural facts rather than free text. Two categories are never included, full stop:
- Break-glass override tokens — never appear in any digest, log, or synced file.
- Response-injection echo snippets — the bounded, DLP-scrubbed excerpt of model output some detector findings carry stays behind the dedicated OWNER/ADMIN-only endpoint it has always lived behind. The log records that a finding was adjudicated and what pattern it matched, never the quoted text.
Turning it on
The Governed Decisions Log is off by default. A growing, auto-written context file is token spend your workspace's agents pay on every request that reads it — that's not something Intutic imposes without an explicit opt-in.
There is no dashboard toggle for it. Enable it through the settings API, with decisionsLogEnabled: true in PUT /api/v1/workspace/settings. Once on:
- The sync-daemon polls a bounded, workspace-scoped projection of the last ~20 decisions on its normal sync cycle.
.intutic/DECISIONS.mdis written locally — the full bounded record, regenerated each cycle. This file is not committed to your repository (it's covered by.gitignore, the same as every other daemon-generated governance artifact).- A short, marker-delimited section (the most recent ~10 entries) is injected into your
CLAUDE.md(or equivalent) — the file your harness already reads as project context — so your coding agent sees recent decisions without needing to know.intutic/DECISIONS.mdexists.
Turning the setting back off stops further updates; it does not delete files a previous sync already wrote.
Refreshing immediately after a merge
If you use Intutic's Git hooks, an optional post-merge hook triggers a one-shot refresh right after git merge, rather than waiting for the daemon's next poll. It's installed alongside Intutic's other hooks and, like all of them, never overwrites a hook that isn't Intutic's own.
Related
- Review Queue — the human-in-the-loop approval flow this log records the outcomes of
- Settings & Config — where to turn this on
- Agent Guidelines (SOPs) — the governance rules your agent enforces, distinct from the decisions log's after-the-fact record