data219

atlcli

Use when the user asks to work with Atlassian Jira or Confluence from the terminal using atlcli, including authentication profiles, Jira issue/sprint/worklog operations, Confluence wiki/docs sync, inline comments, troubleshooting, and JSON output for automation.

data219 3 1 Updated 6mo ago

Resources

4
GitHub

Install

npx skillscat add data219/opencode-docker/atlcli

Install via the SkillsCat registry.

SKILL.md

AtlCLI

Use this skill for Jira/Confluence CLI tasks with atlcli.

Source Of Truth

  • Run atlcli --version first.
  • In v0.14.x, Confluence namespace is atlcli wiki ....
  • For executable command syntax, local atlcli <group> --help is authoritative.
  • Use COMMAND-REFERENCE.md for discovery/routing only.
  • If local help and appendix differ, stop and ask before any mutation.

Quick Routing

  • Issues/search/boards/sprints/worklogs: jira
  • Pages/spaces/comments/docs sync: wiki
  • Auth/profile/config/log/plugin health: root groups (auth, config, doctor, log, plugin)
  • Full command list: COMMAND-REFERENCE.md

Safety Gates (Required)

For any non-read-only command (for example create, update, delete, transition, bulk, push, resolve, assign, comment, attach, link, start/stop, add/remove, import/export, apply, share, set/unset, clear, install/enable/disable):

  1. Run read-first preview when an equivalent read exists (get/list/search --json); otherwise use diff/status/check --strict or a dry-run command.
  2. Scope check: explicit key/ID/filter and impact count.
  3. Use explicit --profile <name>.
  4. High-impact operations always require explicit user ACK in this thread before execution.
  5. Post-write verify via read command and include resulting URL/ID.
  6. If no preview/dry-run/verification path exists, stop and ask for explicit risk acceptance before proceeding.

High-impact definition:

  • Any command with delete, bulk, archive, restore, push, resolve, --fix, update, import, export, install, enable, disable, switch, logout, login, or wildcard-wide filters (for example broad JQL/CQL without tight key/ID scope).

Security Guardrails

  • Never print tokens.
  • Prefer secret managers/CI secrets over inline export ...=<token>.
  • Use least-privilege profiles (separate automation profile per tenant/project/space).
  • After ephemeral env-token use: unset ATLCLI_API_TOKEN.
  • Treat config list and log output as sensitive; redact before sharing.
  • For sensitive or mutating flows, add --no-log where supported.

Inline Comments (Confluence)

atlcli --profile work wiki page comments add-inline --id 12345 --selection "target text" "Please clarify"
atlcli --profile work wiki page comments add-inline --id 12345 --selection "target text" --match-index 2 "Third match"
atlcli --profile work wiki page comments list --id 12345 --inline
atlcli --profile work wiki page comments resolve --comment 67890 --type inline

Rules:

  • Pull/get fresh content before selection-based comments.
  • If selection is repeated, set --match-index intentionally.
  • Use --type inline when resolving/deleting inline threads.

Script Paths

SKILL_DIR="/home/jan/.dotfiles/.config/opencode/skills/atlcli"
$SKILL_DIR/scripts/atl_md_links.sh --help
$SKILL_DIR/scripts/safe_wiki_sync.sh ./docs --profile work

Verification Evidence (RED/GREEN/REFACTOR)

Sources