canonic-canned-loop

name	canonic-canned-loop
description	Day-to-day canned-response loop with canonic: probe free Jira, import drafts, scaffold or promote into corpus/responses, edit, check/lint, reindex/search/dedupe, convert or explicit jira-comment, and rely on CI. Use when curating shared-prefix markdown (canonic.toml), importing Jira comments as review drafts, or publishing one-shot free-tier comments. No bulk auto-sync and no Marketplace apps.

canonic canned-response loop

Markdown under corpus/responses/ is the source of truth. Jira is a publication surface. Free platform REST only (Cloud Free API token or Server/DC PAT).

Constraints (do not break)

• No bulk library push into Jira. jira-comment is one file → one issue.
• No Marketplace / JSM canned-response admin APIs — native REST only.
• Imports are drafts: import-jira writes under corpus/imports/ (or --out), never silently overwrites corpus/responses/.
• Promote only after review: edit drafts, run check, then promote.
• Secrets in gitignored canonic.local.toml under [jira] (never commit tokens).

Config (canonic.toml / canonic.local.toml)

# canonic.toml
prefix = "resp"

# canonic.local.toml (gitignored)
[jira]
base_url = "https://your-instance.atlassian.net"
email = "[e-mail]"
api_token = "..."
# Server/DC PAT alternative:
# auth_header = "Bearer <pat>"

Day-to-day loop

1. Probe free REST

canonic doctor
canonic jira-probe

2. Pull snippets as review drafts

canonic import-jira "project = HSP AND labels = canned-response" --dry-run
canonic import-jira "project = HSP AND labels = canned-response"
# drafts land in corpus/imports/ by default

Fixture smoke (no live Jira):

# optional: scripts/jira-fixture/server.py + [jira] in canonic.local.toml pointing at it
canonic import-jira "project = HSP AND labels = canned-response" --out /tmp/imports

3. Scaffold a new response (template)

canonic new "Project space is not a backup" --tags storage,project-space
# → corpus/responses/resp-project-space-is-not-a-backup.md
canonic new "Queue limits" --id resp-queue-limits --sop none --out corpus/responses
# → corpus/responses/resp-queue-limits.md

4. Edit, then quality gate

# edit the markdown (imports or responses)
canonic check
canonic lint --engine harper          # in-process harper-core; Vale optional
canonic list
canonic tui                           # browse / filter / convert preview (no Jira POST)

5. Promote import → published corpus

canonic promote corpus/imports/resp-some-topic-hsp-101.md
canonic check
canonic reindex
canonic search "project space"
canonic dedupe --reindex --threshold 1.0

6. Render / publish (human-gated)

canonic convert corpus/responses/resp-project-space-is-not-a-backup.md
canonic jira-comment --issue HSP-101 \
  corpus/responses/resp-project-space-is-not-a-backup.md --dry-run
# explicit write only when reviewed:
canonic jira-comment --issue HSP-101 \
  corpus/responses/resp-project-space-is-not-a-backup.md

Cloud Free uses ADF on /rest/api/3; Server/DC uses wiki on /rest/api/2. Override with --body-format wiki|adf when needed.

CI

GitHub Actions runs cargo test --locked --all-targets, then corpus check and in-process Harper lint on corpus/responses/, plus a convert smoke when pandoc is installed. Keep published responses check-clean so CI stays green.

TUI keys (quick)

Key	Action
j/k	Move · / filter · C check · c convert preview
l lint · r reindex · s search · q quit

Full usage: docs/orgmode/usage.org and README Usage section.

Agent checklist

1. Prefer canonic new over hand-copying front matter.
2. Never write import drafts straight into corpus/responses/ without promote.
3. Run check (and lint --engine harper) before convert or jira-comment.
4. Use canonic tui for browse/filter/convert preview — it does not post to Jira.
5. Do not invent bulk sync or paid Jira extensions.
6. Leave production tickets out of git until a human promotes a reviewed draft.