All skills
simota avatar
by shingo imotasimota/agent-skills85 stars
15

Navigating delivery status read-only: reconciles planned scope (specs/roadmap/PRD) against implemented code for what's built vs left. Not for priority scoring (Rank) or AC conformance (Attest).

Use this Skill: https://skilld.dev/gh/simota/agent-skills/pdm

This session only. Nothing lands on disk.

referencesource-triangulation.md

≈1.3k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Source Triangulation Reference

Purpose: Where and how to locate planning artifacts and pair them with code evidence. Read when: You are at LOCATE and need to find the plan side and the code side.

Contents

  • The Two Sides
  • Plan-Side Sources
  • Code-Side Signals
  • Triangulation Strategy
  • Multi-Repo / Multi-Service Scope
  • When a Side Is Missing
  • Source Confidence

The Two Sides

PDM reconciles intent (what was planned) against reality (what is built). LOCATE must establish both. A status produced from one side alone is not a status — it is half a finding.


Plan-Side Sources

Search in rough order of authority. Stop broadening once you have a coherent scope picture.

Source How to find Signal
Specs / PRD / SRS docs/, spec/, *.md with requirement language; request SCRIBE_TO_PDM_HANDOFF Authoritative planned scope
Issue tracker gh issue list --state all (capture state: open/closed/reopened), gh issue list --label feature Granular planned/active work; state is a status signal — see reconciliation.md §Issue-State Cross-Signal
Milestones / Projects gh api repos/:owner/:repo/milestones, project boards Roadmap grouping
Labels gh label list (e.g. epic, mvp, v2) Scope/phase tagging
ROADMAP file ROADMAP.md, roadmap/ Explicit roadmap intent
CHANGELOG / releases CHANGELOG.md, gh release list What was claimed shipped (verify against code)
ADRs docs/adr/, ADR/ Decisions implying planned scope
README "Features"/"Roadmap" README.md sections Lightweight scope when nothing else exists

gh commands require a GitHub remote; fall back to local files when unavailable.


Code-Side Signals

A feature is "built" when code evidence supports it. Signals, strongest first:

Signal How to find Strength
Entry point route/handler/CLI command/exported API for the feature High
Tests describe/it blocks naming the feature behavior High (behavior intended + checked)
Core module a module/dir owning the feature's logic Medium
Feature flag flag config gating the feature Signals In-Progress if off/partial
Stub / TODO empty handler, throw NotImplemented, TODO Signals Not-Started/In-Progress despite presence

Prefer a LENS_TO_PDM_HANDOFF for deep comprehension; PDM's own survey is for lightweight presence checks, not flow tracing.


Triangulation Strategy

1. Plan side first: enumerate planned features/epics from the strongest available source.
2. Code side: for each planned feature, search for entry point → tests → module.
3. Reverse pass: scan code for features with NO matching plan item → Undocumented candidates.
4. Pair each feature: {plan_ref?, code_evidence?} → feed RECONCILE.

The reverse pass (step 3) is what surfaces Undocumented features and is easy to skip — do not.


Multi-Repo / Multi-Service Scope

When a feature spans more than one repo or service (monorepo packages, micro-services, web + BFF + worker), one repo's view is half a status. The Ask First boundary fires when the boundary is unclear — confirm scope first, then reconcile per the rules below.

Confirm before reconciling: which repos/services are in scope, and whether the question is per-service or end-to-end (a feature can be Done in the API yet Not-Started in the UI).

Situation Strategy
Monorepo, multiple packages Use the package/dir as the area; the plan side is usually shared (one issue tracker). Reconcile each package, roll up per feature across packages.
Multi-service, one feature crosses services Reconcile the feature per service, then derive an end-to-end status: it is Done only when every required service is Done; otherwise In-Progress with the lagging service named.
Separate repos, separate trackers Locate each repo's plan side independently; never assume one repo's roadmap covers another. Flag cross-repo gaps as drift.
Scope boundary unclear Ask First — do not silently pick one repo and present it as the whole.

End-to-end rollup rule: a cross-service feature's status is the weakest of its per-service statuses, and its confidence the lowest of its parts. Name which service holds it back so the gap is actionable.


When a Side Is Missing

Situation Action
No plan source at all Ask First: infer scope from code + README only (cap confidence at Medium) or stop.
No issue tracker, specs exist Use specs; note issues unavailable.
CHANGELOG claims shipped, no code Treat as drift — verify, flag, do not trust the claim.
Code present, no plan ref Undocumented; do not invent a spec for it.

Source Confidence

Record where each side came from; it sets the ceiling on status confidence:

  • Spec text read + code read → High eligible.
  • Issue title only + entry point only → Medium.
  • README bullet + no code, or code + no plan → Low (and usually Not-Started/Undocumented).

Source: SKILL.md on GitHub

No alerts13d3 checks · Risk SAFE
  • Gen Agent Trust Hub13d

    The pdm skill is a project management and navigation tool that reconciles documentation with code implementation. It is classified as low risk because it ingests untrusted data from issue trackers and project specifications, which is a known vector for indirect prompt injection.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

Signed by skilld at 35ffd55. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/pdm