goal-backlog
Skill509 starsChanged 3 months agoMITView on GitHub
Groom the open task pool goal by goal: see everything under each quarter goal, confirm doubtful goal links, retire what has gone stale, and mark what gets picked up first when that goal earns week-time. Use when the user says 'groom my backlog', 'what's under this goal', 'my tasks are a swamp', 'clean up my tasks', or when open tasks have piled up untouched. Not for routing new inbox items; use `triage`. Not for setting the week's priorities; use `week-plan`.
In this repository153 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .claude/plugins/compound-engineering/CLAUDE.mdCLAUDE.md
- .cursor/rules/search-routing.mdcCursor rule
- anthropic-algorithmic-artSkill
- anthropic-brand-guidelinesSkill
- anthropic-doc-coauthoringSkill
- anthropic-docxSkill
- anthropic-frontend-designSkill
- anthropic-internal-commsSkill
- anthropic-mcp-builderSkill
- anthropic-pptxSkill
- anthropic-slack-gif-creatorSkill
- anthropic-theme-factorySkill
- anthropic-webapp-testingSkill
- anthropic-web-artifacts-builderSkill
- anthropic-xlsxSkill
- atlassian-setupSkill
- backup-nowSkill
- backup-restoreSkill
- backup-setupSkill
- calendar-setupSkill
- commitmentsSkill
- create-mcpSkill
- daily-planSkill
- daily-reviewSkill
- decision-logSkill
- delegate-checkSkill
- dex-backlogSkill
- dex-doctorSkill
- dex-improveSkill
- dex-obsidian-setupSkill
- dex-rollbackSkill
- dex-updateSkill
- dex-whats-newSkill
- diff-profileSkill
- enable-semantic-searchSkill
- getting-startedSkill
- google-workspace-setupSkill
- granola-setupSkill
- identity-snapshotSkill
- industry-truthsSkill
- initiative-kickoffSkill
- integrate-mcpSkill
- journalSkill
- manage-capabilitiesSkill
- meeting-closeoutSkill
- meeting-prepSkill
- ms-teams-setupSkill
- pipedrive-setupSkill
- pipeline-syncSkill
- process-meetingsSkill
- product-briefSkill
- project-healthSkill
- prompt-improverSkill
- relationship-radarSkill
- reviewSkill
- save-insightSkill
- scrapeSkill
- skill-scoreSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 4 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 23
Call get_goal_backlog(goal_id="all") (or the single goal ID the user named).
- Very specific: line 30
If the pool is empty (open_task_count is 0), say so plainly and stop — no invented work.
- Distinctive: line 61
Ask the deciding question plainly: "When this goal gets week-time, what should be picked up first?"
- Distinctive: line 67
After the goal groups, walk the pillar groups, then the fully orphaned tasks:
- Distinctive: line 81
If declined, drop it — don't raise it again this session, and don't repeat the offer per goal.
- Distinctive: line 105
Nagging about provisional goals: after one declined offer.
- Very specific: line 111
Update System/usage_log.md to mark goal-backlog grooming as used.
- Very specific: line 113
Analytics (Silent): Call track_event with event_name goal_backlog_completed and properties goals_groomed, tasks_retired, next_up_set.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- Generated from `.claude/skills/goal-backlog/SKILL.md` by `scripts/generate-agents-skills.py`. Do not edit. -->Turn a swampy task list back into a groomed pool: one goal at a time, decide what's still real, what's next, and what can go — so /week-plan pulls from decisions you already made instead of a flat dump.
Usage
/goal-backlog— Walk every group, one goal at a time/goal-backlog Q3-2026-goal-2— Groom one goal only
The one rule that overrides everything
Never delete or move a task silently. Every exit from the backlog is the user's decision, and before any line is removed or relocated you show the exact lines that will change. Completing a task through the Work MCP is the only exit that needs no preview, because nothing is removed.
Step 1: Load the backlog
Call get_goal_backlog(goal_id="all") (or the single goal ID the user named).
Handle the response per CLAUDE.md's feature_status convention:
off(Quarter Goals room off): one calm line with the returneduser_message, no error tone, and stop — there is nothing to groom without goals.not_installed/broken: surface the returneduser_messageand its fix path, then stop.unknown, or the tool errors: say the backlog could not be checked. Do not improvise a grooming pass from grep.
If the pool is empty (open_task_count is 0), say so plainly and stop — no invented work.
Then show a one-screen overview in plain language, largest and stalest groups first:
"27 open tasks. Goal 2 (Launch mobile beta) holds 14 — 9 untouched for 3+ weeks. Goal 1 holds 6, all fresh. 5 tasks serve a pillar but no goal, and 2 are linked to nothing at all. Where do you want to start?"
Staleness comes from each task's staleness_days (age since the task was created); a group's stale_count uses the 3-week line the tool reports. Say "sat untouched for 3+ weeks" — never dress the number up as activity data it isn't.
Step 2: Groom one goal at a time
For the chosen group, show: the count, the stale flags, tentative (?) links, and the current next-up order if one exists. Keep it to one screen; titles and ages, not raw metadata.
Work through three passes, in this order:
2a. Settle the doubtful links
For each task marked tentative, ask one short question — "Does 'Draft partner FAQ' really serve this goal?" — and call confirm_goal_link(task_id, action="confirm") or action="clear". Cleared tasks fall back to their pillar group; mention that so they don't seem to vanish.
2b. Retire what's dead — completion is not the only exit
For stale items (and anything the user calls out), offer four honest exits:
| Exit | What happens | How |
| Done | It was actually finished — mark it complete everywhere | update_task_status(task_id, status="d") |
| Someday | Still real, not this quarter — park it out of the open pool | Move the task line and its indented child bullets from 03-Tasks/Tasks.md to 03-Tasks/Someday.md (create that file with a one-line header if missing). Show the exact lines before moving. Nothing is deleted; moving the lines back revives the task. |
| Delete | It's noise — remove it for good | Show the exact lines that will be removed and get an explicit yes for that task first. Never batch-delete on one blanket approval. |
| Keep | Still earning its place | No change |
2c. Set the next-up order for the survivors
Ask the deciding question plainly: "When this goal gets week-time, what should be picked up first?"
For each item the user sequences, call set_task_next_up(task_id, next_up=N) (1 = first). To drop something out of the order, set_task_next_up(task_id, next_up=null). Suggest an order from priority and age if asked, but the user ranks — don't renumber a queue they already set without asking.
Step 3: Pillar-only and orphaned work — last
After the goal groups, walk the pillar groups, then the fully orphaned tasks:
- For each, offer to link it to a goal — or consciously leave it operational; not everything must ladder up, and saying so is a fine outcome.
- To add a link, append the documented indented child bullet under the task line in
03-Tasks/Tasks.md— for example- Goal: Q3-2026-goal-2— matching the metadata format Dex writes itself. Show the edit before making it. - Orphans with no pillar either get a pillar bullet the same way or stay put by choice.
Pillar groups can carry a next-up order too — same question, same tool.
Step 4: Provisional goals — offer once, then move on
A group flagged provisional was recovered from freeform text: its ID is generated and tasks cannot link to it until it is structured. Offer once:
"Two of your goals were recovered from freeform notes. Want me to structure them into the quarter goals page shape (
### N. Title — **Pillar** ^Qn-YYYY-goal-N) so tasks can link to them?"
If declined, drop it — don't raise it again this session, and don't repeat the offer per goal.
Step 5: Close with a one-screen summary
End with exactly what changed, nothing rhetorical:
"Grooming done.
- Goal 2: confirmed 2 links, cleared 1, marked 1 done, parked 3 in Someday, next-up order set (3 items).
- Goal 1: untouched — all fresh.
- Linked 2 pillar tasks to Goal 3; left 3 operational by your call.
- Deleted 1 task (you approved the exact line).
/week-planwill now pull each goal's next-up items first. To undo an order, ask me to clear next-up on a task; to revive a parked task, move its lines back from03-Tasks/Someday.md."
Quality bar
A good grooming pass leaves every surviving task deliberately kept, every stale item decided (not re-snoozed by silence), and at least the most active goals with a next-up order /week-plan can pull from. The user should be able to say afterwards what left the pool and why.
Anti-patterns
- Deleting or moving anything without showing the exact lines and getting a yes. This is the cardinal sin of a groomer.
- Marking a task done to tidy it away. Done means finished; use Someday or delete for the rest.
- Nagging about provisional goals after one declined offer.
- Renumbering the user's existing next-up order on your own initiative.
- Grooming into planning. Choosing the week's priorities is
/week-plan's job; hand off there instead of duplicating it.
Track Usage (Silent)
Update System/usage_log.md to mark goal-backlog grooming as used.
Analytics (Silent): Call track_event with event_name goal_backlog_completed and properties goals_groomed, tasks_retired, next_up_set. Fires only if the user opted into analytics; no action if it returns "analytics_disabled".
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

