agentleFS
nesnilnehcnesnilnehc/ai-cortex/skills/update-roadmap/SKILL.md

update-roadmap

Skill8 starsChanged 2 months agoMITView on GitHub

Maintain requirement-and-defect roadmap summaries — change item status, shift dates with downstream impact analysis, and retain source-backed revision history. Does not move items between Now/Next/Later tiers.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

20 rules. About 0 are common, 15 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈15

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the existing structure is adequate, use this skill only for separately authorized source-backed maintenance.

    line 19
  2. Distinctive:

    Acceptance test: after the change, can someone who took no part in the discussion read the summary alone and state "what changed, why, and …

    line 39
  3. Very specific:

    Handoff point: if the change really calls for a cross-tier adjustment (an item that no longer belongs in Now, say), stop and hand off to …

    line 41
  4. Distinctive:

    Default: read roadmap.md and the related item files from the paths the project norms define

    line 79
  5. Distinctive:

    Never add a task or task group as a roadmap row.

    line 88
  6. Very specific:

    If what the user actually wants is a move across tiers → stop, explain that this falls to promote-roadmap-items, and hand off.

    line 90
  7. Distinctive:

    Verify the source basis for every transition, including pause, resume and reopening.

    line 94
  8. Distinctive:

    Before Completed, verify the project's declared closure criterion and linked acceptance and production-release evidence where required.

    line 95
  9. Distinctive:

    Ask for the reason for the shift (scope change / dependency slip / resource change / other)

    line 98
  10. Very specific:

    Compute the downstream impact: read each item's depends_on, find the items that carry this one as a prerequisite, and list their affected …

    line 99
  11. Very specific:

    Update the authoritative source and, when the parent planning summary changes, its roadmap row.

    line 114
  12. Distinctive:

    Do not change the structure: no new milestones, no changing stage promotion criteria, no editing the strategic bets

    line 170
  13. Distinctive:

    Ask the two questions: what it is stuck on (the third-party sandbox is not open yet) and what the plan is (ticket filed, 3 working days …

    line 221
  14. Very specific:

    Read depends_on and find that #51 advanced reporting carries #42 as a prerequisite → downstream impact: #51's start date moves out.

    line 222
  15. Distinctive:

    Check the history: this item has already shifted twice within this cycle.

    line 236
  16. Distinctive:

    Before a material planning change, read the current roadmap and durable history.

    line 275
  17. Distinctive:

    Read linked delivery/history pages and declared authoritative records before edits.

    line 292
  18. Distinctive:

    Read the existing selection source and reuse primary perspective, views, layout and destination.

    line 306
  19. Distinctive:

    Before writing roadmap-chain deliverables, apply the language contract.

    line 319
  20. Distinctive:

    Read the user instruction, consuming-project policy and existing artifact first.

    line 320

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Skill: Update Roadmap

Purpose

The entry point for day-to-day maintenance once the roadmap is in place: change status, shift dates, say plainly what this round changed.

define-roadmap covers building from scratch, promote-roadmap-items covers promotion and demotion across tiers, and between them sits a large block of day-to-day movement — an item started, an item got stuck, an item slips by two weeks — which is what this skill covers.

First alignment of an existing project's nonconforming structure goes to define-roadmap under the Spec's preservation contract. If the existing structure is adequate, use this skill only for separately authorized source-backed maintenance; alignment is not permission to change dates, membership or execution. Mixed requests separate structural repair from the stated business change, retain one source/revision identity where applicable, and re-review the result.


Core Objective

Primary goal: land one roadmap change completely, and let a reader see what changed, why it changed, and who it affects.

Success criteria (all must hold):

  1. ✅ A requirement or defect status change agrees with its source and roadmap summary; task-only changes stay in the task list unless they affect the parent outcome or key blocker
  2. ✅ A risk escalation or a change to Blocked must record the blocking reason and the mitigation; with either one missing, nothing is written
  3. ✅ A date shift must compute the downstream impact: list the items this shift affects
  4. ✅ Items that cross a hard deadline after the shift are flagged explicitly
  5. ✅ A "what changed this round" summary is output, covering the changed items, the reasons, and the blast radius
  6. ✅ The Now / Next / Later tier each item sits in stays unchanged

Acceptance test: after the change, can someone who took no part in the discussion read the summary alone and state "what changed, why, and who is affected"?

Handoff point: if the change really calls for a cross-tier adjustment (an item that no longer belongs in Now, say), stop and hand off to promote-roadmap-items.


Scope Boundaries

This skill owns:

  • Execution status changes using the fixed vocabulary in roadmap-quality; risk remains a separate source annotation
  • Date and timing adjustments, and the downstream impact analysis that goes with them
  • Moving requirements/defects into or out of existing versions, preserving tier
  • The what-changed summary for this round

This skill does not own:

  • Cross-tier promotion / demotion (promote-roadmap-items)
  • Structural roadmap changes: new milestones, stage promotion criteria edits, strategic bet edits (define-roadmap)
  • Creating items (capture-work-items)
  • Re-scoring (prioritize-backlog)
  • Dependency identification (map-item-dependencies)

The line against promote-roadmap-items: changing the tier belongs to promote, leaving the tier alone belongs here. Pushing a Now item back two weeks is this skill; moving it to Next is promote. The test is whether the tier changed, not whether the date changed.


Use Cases

  • Progress sync: an item started / finished and the status has to keep up
  • Risk escalation: an item is stuck, and the blocking reason and mitigation need recording
  • Timing adjustment: a dependency slipped or the scope changed, so the date moves back and the blast radius has to be visible
  • Before communicating a change: a "what changed this round" note is needed to share with stakeholders

Behavior

Interaction policy

  • Default: read roadmap.md and the related item files from the paths the project norms define
  • Must ask: on a risk escalation or a change to Blocked, the blocking reason and the mitigation are both indispensable; if either is missing, nothing is written
  • Write the requested change: analyze downstream impact before persisting. Ask only when the target, intended state, or a hard-deadline decision remains unclear.

Execution

  1. Read the current state: discover the established local or remote authority from project norms; load its latest roadmap and affected details, roadmap-quality, project closure criteria and source requirement or defect records. Do not revive a retired local draft when the remote page is authoritative. Resolve task-level input to its parent. Update task details in their source; only refresh the parent roadmap summary if its outcome, status, timing or key blocker changes. Never add a task or task group as a roadmap row.
  2. Identify the change type: status change / date shift / existing-version scope change / existing-iteration goal, scope or cadence change / combination.
    • If what the user actually wants is a move across tiers → stop, explain that this falls to promote-roadmap-items, and hand off.
  3. Handling a status change:
    • Map the project's source execution evidence to exactly one label from the quality rule. Keep approval, execution and key conditions in separate columns. Risk alone does not replace an execution label.
    • For a risk escalation or Blocked change, obtain the reason and mitigation from sources first; ask only for missing facts. Preserve the existing requirement to have both before writing. Keep details in the source and only the short next-step condition and link in the roadmap.
    • Verify the source basis for every transition, including pause, resume and reopening. Development complete maps to Awaiting acceptance while required acceptance is pending; passed acceptance maps to Awaiting release only if release remains required for closure. A recorded pause or blocker takes precedence while applicable.
    • Before Completed, verify the project's declared closure criterion and linked acceptance and production-release evidence where required. Do not invent or waive gates. If closure or the mapping is undefined, or sources conflict, report the limitation outside the status cell and leave the transition unresolved.
    • Task counts, test totals, commits and approval alone never prove completion; use metrics only when source criteria define them.
  4. Handling a date shift:
    • Ask for the reason for the shift (scope change / dependency slip / resource change / other)
    • Compute the downstream impact: read each item's depends_on, find the items that carry this one as a prerequisite, and list their affected dates one by one
    • Flag hard-deadline breaches: items that cross a compliance, commitment, external-constraint or similar hard deadline after the shift are marked in red on their own
    • With the dependency data missing (map-item-dependencies was never run), state outright "downstream impact not computed, dependency data missing"; do not pretend the analysis happened
  5. Review the change list: changed items + reasons + downstream impact + deadline breaches. Resolve any ambiguity or hard-deadline decision with the user before applying it.
  6. Persist source and summary: load roadmap modeling. Retain the single policy authority and refresh the current arrangement explanation only from settled decisions; maintenance does not introduce another priority rule. Preserve the reader hierarchy with material unknowns early and policy/closure/ selection/language evidence later. Do not create or rename related pages merely to adopt suggested document-family names; existing titles/paths/URLs remain valid. For a move between existing versions, record old/new membership and the effect on both delivery windows, dependencies and commitments; leave tier unchanged. New versions/iterations, layout changes or structural changes go to define-roadmap. Refresh all selected views in the established destination using that skill's generator procedure when useful; confirm only a changed destination, not the same authorized location again. Update the authoritative source and, when the parent planning summary changes, its roadmap row. Keep test counts, task IDs and detailed evidence in source records; link them instead. Source/summary/diagram disagreement remains unresolved until repaired.
  7. Output the change summary: see the template below.

For a remote destination, follow the platform delivery guide and its destination-specific adapter. Re-read before writing and reconcile concurrent edits. Refresh affected summaries/charts from their declared authority, then verify read-back, protected links/anchors/cards/history and affected reader behavior. Reuse still-valid evidence for unaffected content/assets; a metadata-only edit must preserve the exact rich body. Source/diagram changes require fresh affected checks and cannot hide behind that narrower scope. A formatting-only change needs no fabricated business revision; use remote revision history where Git does not own the document. Clean obsolete publication wording only in living text. Report partial delivery and unverified source facts without claiming an automatic quality guarantee or changing the source system outside authorization.

Change summary template

## Roadmap changes this round

**Changed at**: <ISO date>

| Item | Change | Reason |
| --- | --- | --- |
| #42 payment optimization | In progress → Blocked | waiting on the third-party sandbox |
| #63 mobile workflow | date +2 weeks | design delivery slipped |

**Blocked items and mitigations**
- #42: ticket filed, 3 working days expected; a local stand-in is being prepared so development can go on in parallel

**Downstream impact**
- #51 advanced reporting: prerequisite #42 slipped, so this item's real start date moves out 2 weeks

**Hard-deadline breaches**
- None

**Unchanged**: the tiers stay as they are (this skill makes no cross-tier adjustment)

Input & Output

Input: the current roadmap.md, the intended change, and optional blocker details.

Output: updated authoritative source and, where parent planning changed, its roadmap summary; downstream impact and change summary.


Restrictions

Hard Boundaries

  • Do not change the tier: an item must not be moved between Now / Next / Later; that is promote-roadmap-items's job
  • Do not change the structure: no new milestones, no changing stage promotion criteria, no editing the strategic bets
  • A blocker must have a follow-up: a risk escalation / Blocked change is not written while the blocking reason or the mitigation is missing
  • Source agreement: requirement and defect summaries agree with their source; task-only progress need not change the roadmap
  • With the dependency data missing, must not pretend the downstream impact was analyzed — the fact that it was not computed, and why, must be stated explicitly

Anti-patterns (avoid)

  • ❌ Changing status silently: change it without saying why, and next time nobody remembers what happened
  • ❌ Shifting a date without looking downstream: pushing one item back looks harmless and drags a whole chain with it
  • ❌ Using a date shift in place of demotion: an item shifted over and over does not belong in Now; the route is demotion via promote-roadmap-items, not one more push
  • ❌ Marking done without acceptance evidence: task progress or approval does not prove the requirement or defect outcome
  • ❌ Changing a pile of things with no summary: the change summary is what stakeholders read, and a batch of changes needs it more, not less

Skill Boundaries (avoid overlap)

ActionOwner
Cross-tier promotion / demotionpromote-roadmap-items
New milestone / stage criteria editdefine-roadmap
Creating itemscapture-work-items
Re-scoringprioritize-backlog
Dependency identificationmap-item-dependencies
Roadmap health checkreview-roadmap

Self-Check

  • The change type was identified; anything that is a cross-tier move went to promote-roadmap-items
  • Every risk escalation / Blocked change recorded the blocking reason and the mitigation
  • Every transition has source evidence; Completed meets declared closure including acceptance/release where required; no gates were invented
  • Date shifts had their downstream impact computed; where dependency data was missing, the omission was stated explicitly
  • Hard-deadline breaches were flagged on their own
  • The requested change and downstream impact were reviewed before persistence; any unresolved decision was clarified
  • Separate approval, execution and key conditions columns follow the quality rule; each execution cell contains one fixed label, without risk labels or evidence
  • Source and parent summary agree; task-only changes did not add roadmap entries or duplicate details
  • The what-changed summary was output
  • No item's tier was touched
  • Existing-version scope changes preserve tier and report both versions' impact; all selected views agree.

Examples

Example 1: an item gets stuck (mainstream case)

Background: #42 payment optimization sits in the Now tier, in progress, and is suddenly stuck on a third-party sandbox.

Flow:

  1. Identified as a status change: In progress → Blocked.
  2. Ask the two questions: what it is stuck on (the third-party sandbox is not open yet) and what the plan is (ticket filed, 3 working days expected; a local stand-in is being prepared for parallel development).
  3. Read depends_on and find that #51 advanced reporting carries #42 as a prerequisite → downstream impact: #51's start date moves out.
  4. No hard-deadline breach.
  5. The target and impact are clear → roadmap.md and #42's frontmatter are both written.
  6. Output the change summary.

Result: the blocker is on record with a mitigation, and #51's slip surfaces early rather than at the moment it was due to start.

Example 2: repeated shifts that call for demotion (edge case)

Background: #63 mobile workflow is being pushed back for the third time.

Flow:

  1. Identified as a date shift; the reason asked for → "design never gets scheduled".
  2. Check the history: this item has already shifted twice within this cycle.
  3. Tell the user: an item that shifts over and over in the Now tier does not currently meet the conditions for being pulled into Now — shifting it again only leaves the Now tier looking full while nothing moves.
  4. Suggest switching to promote-roadmap-items and demoting it to Next, keeping the Now tier focused on items that can move.
  5. The user agrees → this skill stops and hands off to promote.

Result: a third shift did not paper over the problem; the tier adjustment went back to the skill it belongs to.

Example 3: dependency data missing (edge case)

Background: the user wants to push #38 data pipeline upgrade back by 3 weeks, but the project has never run map-item-dependencies and the items carry no depends_on.

Flow:

  1. Identified as a date shift; the reason is that resources were pulled away.
  2. Attempt to compute the downstream impact → every item's depends_on is empty rather than —, so "no dependency" cannot be told apart from "not checked".
  3. No pretending the analysis happened. The change summary says outright: "downstream impact not computed — the project has not registered dependency data yet".
  4. Suggest running map-item-dependencies first and then deciding how far the shift goes, along with a pointer the user can judge from: pipeline items usually have downstream consumers.
  5. The user chooses to shift now and fill in the dependencies later → persist as decided, keeping the not-computed statement in the summary.

Result: the change goes ahead, but "downstream was not computed this time" is written out in the open rather than leaving a reader to assume it was assessed.

Filled deliverable example

Filled change record and adjusted views. This is a fictional completed output, not a template or authorization to change the adopting project.

Phase and milestone contract

Maintain existing phase timing and milestone timing/state from linked source evidence. Apply the Roadmap Spec: actual event timing requires reached state; non-planned states require evidence. Date passage, member completion and scope changes do not automatically reach a milestone or approve related work. Report affected version windows, phase spans and event targets separately. Route new phase/event structure to define-roadmap; do not silently upgrade legacy inputs.

Self-check: concepts, selected views and linked evidence agree with the source records.

Revision history maintenance

Before a material planning change, read the current roadmap and durable history. Apply the revision contract: append one source-backed decision with a stable ID, date, before/after, reason, impact and details link. Refresh settled priorities from their authoritative source without making priority decisions here. Record priority/tier handoffs, membership, phase, timing/commitment, exclusion and milestone-state changes where applicable. The conversational change summary is also retained in the project history source; it is not the only copy. Keep the latest five in the roadmap and verify the full history link before shortening it. Re-rendering or retrying the same decision reuses its ID; a correction appends a linked revision. Formatting-only edits use the owning platform's revision history (Git for repository documents). Missing decision evidence stays unresolved, never invented.

Self-check: material changes are traceable, retained links resolve, and repeated execution adds no duplicate revision.

Iteration and document maintenance

Read linked delivery/history pages and declared authoritative records before edits. Maintain existing iteration goals, scope, cadence and timeboxes only from authorized source decisions. Analyze affected versions, dependencies and commitments; carry-over alone changes no tier, closure state or release evidence. New iteration structure or layout routes to define-roadmap. Keep execution-board tasks in their source. Refresh all affected main/detail/chart summaries together, reusing the same revision ID. Preserve source-relative links and selected single/split layout. Report any projection that could not be updated; do not claim a consistent plan while it differs.

Self-check: iteration changes have source/revision evidence; every affected main, detail and time view agrees, with no duplicate editable scope or dates.

Retained scenario selection

Read the existing selection source and reuse primary perspective, views, layout and destination. Missing new metadata in legacy plans is not a reason to re-interview or migrate. Update flow summaries from linked policy/board evidence when applicable; no fabricated cadence, WIP number or service target. If needs materially change, report the difference and route a structural reconsideration to define-roadmap while preserving current choices. Refresh every affected projection only within the settled combination; do not apply a fresh recommendation during status/date maintenance.

Self-check: the update preserves the selection or links the authorized new decision; changed context and unavailable evidence are explicit.

Deliverable language

Before writing roadmap-chain deliverables, apply the language contract. Read the user instruction, consuming-project policy and existing artifact first; retain a settled choice across handoffs. For a new artifact infer only from consistent reader/document evidence; ask only unresolved gaps or winning-level conflicts. Conversation language and English templates alone do not select artifact language. Record the choice and source in existing norms/planning decisions; no new mandatory configuration. Localize all new prose and display labels consistently while preserving IDs, machine fields, canonical enums, source links and historical originals.

Apply the retained choice to new reports, rationale, tables and diagram labels as well as refreshed roadmap projections. Preserve existing source-record policies; this workflow does not authorize bulk translation of unrelated documents.

Self-check: language evidence and retained choice are explicit; all new display prose is consistent, machine contracts are unchanged, and unavailable checks are reported.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.