# confluence-doc canary fixtures

Planted doc-gap corpus for `checkers/confluence-doc.sh --canary` (offline, no network/token).
The checker asserts the total doc-gap count equals `EXPECTED_GAP_COUNT` (anti-complacency
floor, design §6.4). If a gap check regresses (stops firing) or the fixture changes, the count
drifts and the canary FAILS (exit 3).

`--canary` implies `--dry-run + --no-api`, so the LIVE Confluence API checks (page-existence +
staleness, which need the gated `confluence-bot` token, D6) are SKIPPED and noted — they are
never counted as a gap on missing data (memory `feedback_cloudwatch_alarms`).

## Fixture inputs

| File | Role |
|---|---|
| `repos.txt` | the repo set to diff against the page-ID map (one repo name per line) |
| `mock-page-map.json` | a MOCK IT page-ID map (same shape as `project_confluence_migration`) |
| `mock-aws-inventory.json` | a MOCK read-only AWS inventory (what the API/collector would return) |

## The 3 planted gaps

| Check | Subject | Why it's a gap |
|---|---|---|
| repo-documented | `orphan-tool-repo` | no page in the mock map (and not doc-exempt) |
| aws-documented | `afi-backup-monitor` (Lambda) | inventory resource with no page in the mock map |
| required-page | `IAM & Access Management` | a REQUIRED standing page omitted from the mock map |

Non-gaps proving the checks are precise (must NOT inflate the count):
- `payments-dashboard`, `seahaven-slack-bot` repos → matched to their pages.
- `engineering-handbook` repo → `DOC_EXEMPT_REPOS` → skipped, not a gap.
- `payments-dashboard` Lambda → matched to the "Payments Dashboard" page.
- `Incident Response Runbooks`, `Backup & Disaster Recovery` required pages → present in the map.
- The LIVE API staleness/existence check → SKIPPED (no creds in canary), noted, not a gap.

Total = **3** (`EXPECTED_GAP_COUNT`).

When you add/remove a check, a fixture input, or a planted gap, update the fixture(s) and
`EXPECTED_GAP_COUNT` in the same commit (the canary edit is itself caught on the next run —
design §6.4).

## Not exercised offline (PROVISIONING — gated)

The LIVE Confluence reads (and the on-demand WRITE path via
`~/.claude/scripts/confluence_mermaid.py`, including the page-1540098 live dry-run that must list
all 16 weweave Mermaid macros) require the `confluence-bot` service account + token. That account
creation, its 90-day rotation, and the Mermaid live dry-run are provisioning steps documented in
the checker's PROVISIONING footer — they are NOT performed by the canary.