issue-estimate-sp
Estimate story points for a Jira ticket using historical data and multi-persona scrum poker. Use when user says "estimate this ticket", "how many points", "story point estimate", "estimate [TICKET-KEY
SKILL.mdskills/issue-estimate-sp/SKILL.md
GitHubEstimate
Estimate story points for a Jira ticket via context gathering and a scrum-poker council session. This is a read-only operation. Never update the ticket.
Input
ticket_key(required) — e.g.ABC-12345
Workflow
Estimate Progress:
- [ ] Step 1: Gather ticket context from Jira
- [ ] Step 2: Gather related context (feature, linked tickets, Confluence)
- [ ] Step 3: Load reference data for calibration
- [ ] Step 4: Run scrum-poker council session
- [ ] Step 5: Present estimate with rationale
Step 1: Gather ticket context
Use the Atlassian MCP (read-only) to collect:
- Issue details —
jira_get_issuewith the ticket key. Extract: summary, description, issue type, priority, labels, components, current story points (if any), acceptance criteria. - Comments —
jira_get_commentsfor team discussion and context. - Changelog —
jira_get_changelogto see if story points were previously set, changed, or debated.
Step 2: Gather related context
- Parent feature/epic — if the issue has a Feature Link or Epic Link, fetch it with
jira_get_issueto understand the broader scope. - Linked tickets — follow issue links (blocks, relates-to, split-from) to understand dependencies and prior work. Limit to 3-5 most relevant links.
- Confluence context — if the ticket references Confluence pages or has remote links, fetch up to 2 pages for additional design/requirements context using
confluence_get_page. - Similar completed tickets — search for recently closed tickets with similar labels (components are rarely populated — only 8% of tickets have them). Use team labels (
emerald,pyrite,obsidian) and category labels (POAM,sp-candidate,spTooling,documentation) for matching. Example JQL:project = {PROJECT} AND labels in ({TEAM_LABEL}) AND status in (Done) AND "Story Points" is not EMPTY ORDER BY resolutiondate DESC(limit 5). Also check for keyword-similar tickets using a text search on the summary.
Step 3: Run scrum-poker council session
Invoke the clarity-council skill with the following configuration:
user_problem: "Estimate story points for {TICKET_KEY}: {SUMMARY}"
context: |
{Paste the gathered ticket description, acceptance criteria, and key details}
Related context:
{Feature/epic summary if available}
{Key linked ticket summaries}
Reference calibration:
{Summary of relevant reference data from REFERENCE_DATA.md —
include the representative tickets at the 2-3 most likely SP levels}
desired_outcome: >
A Fibonacci story point estimate (1, 2, 3, 5, 8, 13, or 21) with rationale.
Each persona should independently "play a card" (pick a number),
then the group synthesizes to a final recommendation.
constraints:
- 1 story point ≈ 1 business day of flow time (In Progress to Done)
- Use only Fibonacci values: 1, 2, 3, 5, 8, 13, 21
- Account for coordination overhead, not just coding effort
- Investigation/prototype tickets tend to resolve faster than estimated
- Consider review queue time as part of the estimate
- This is READ ONLY — do not modify the ticket
selected_personas:
- scrum-master
- tech-lead
- senior-developer
- qa-engineer
depth: standard
Persona overrides for scrum poker:
- scrum-master: Focus on historical velocity, team capacity, and whether this fits cleanly into a sprint. Compare against reference data for similar tickets. Watch for estimation anti-patterns (artificial splitting, optimism bias).
- tech-lead: Focus on technical complexity, unknowns, dependencies, and coordination tax. Flag if this crosses team boundaries or requires environment access.
- senior-developer: Focus on implementation effort — lines of code, number of files, testing complexity, review burden. Be concrete about what "done" looks like.
- qa-engineer: Focus on test coverage needed, edge cases, regression risk, and validation effort. Flag if testing requires a deployed environment or special access.
Step 4: Present estimate
Format the output as:
## Story Point Estimate: {TICKET_KEY}
**Recommended: {N} story points**
### Scrum Poker Results
| Persona | Card Played | Key Reasoning |
|---------|-------------|---------------|
| Scrum Master | {N} | {one sentence} |
| Tech Lead | {N} | {one sentence} |
| Senior Developer | {N} | {one sentence} |
| QA Engineer | {N} | {one sentence} |
### Why {N} points
{2-3 sentences synthesizing the council's reasoning. Reference the closest
comparable ticket from REFERENCE_DATA.md if one exists.}
### Risk factors
- {bullet list of things that could push the estimate higher}
### Comparable tickets
| Ticket | Summary | Est. SP | Actual Days |
|--------|---------|---------|-------------|
| {key} | {summary} | {sp} | {days} |
> **Note:** This estimate was generated from historical data and persona analysis.
> It has not been written to the ticket. Discuss with your team before committing.
Quality Loop
Before returning, check the output against these criteria. If two or more fail, revise and re-check — at most two passes, then note what still falls short.
- Estimate cites comparable historical tickets, not just a bare number
- The scrum-poker spread (low/mode/high) is shown, not only the final point value
- Key uncertainty drivers that would move the estimate are named
- Read-only respected — the ticket is not modified
<!-- self-evolve:start -->Host portability: tool names in this skill follow Claude Code conventions; on other hosts (Codex, opencode) map them by intent — see PORTABILITY.md.
Self-Evolve Loop
Journal ~/.ink-and-agency/learnings/issue-estimate-sp.md (or workspace-local .ink-and-agency/ where the sandbox confines writes). Read it first; append what the run taught — SELF-EVOLVE.md.