SuperAPI AI Directory

project-status

Coding & development

View repository

Updated 10-16 05:43

About this skill

Project Status

Produce a concise, evidence-backed briefing and recommend one next action. Keep the run read-only: inspect local and remote state without editing code, documentation, planning state, git state, or GitHub state. End after the briefing; implementation requires a separate request.

Treat triangulation as the governing rule. Plans, roadmaps, TODOs, handoffs, and status documents describe intent or remembered state; verify their material claims against independent repository evidence before presenting them as current reality.

1. Establish the Snapshot

  1. Read applicable instruction files and honor any user-named scope. Use the current conversation to locate active work, then verify its project-state claims by the same triangulation rule. Default to the current repository and worktree; include other worktrees or repositories only when the user requests an aggregate view.
  2. Capture the local baseline before relying on status documents:
    • current branch, upstream, and ahead/behind state;
    • staged, unstaged, and untracked work, summarized by purpose when the diff makes that purpose evident;
    • a small recent history window, widened only when it cannot explain the apparent current work.
  3. Discover the repository's compact routing and status sources. Prefer user-named files, then applicable README or architecture routing, roadmaps, plans, task lists, progress files, and handoffs. Follow only links that could determine current work, recent completion, blockers, or the next action.

This phase is complete when the reported current and next work, recent movement, local in-progress work, and relevant scope are known as claims to verify. Absence of a planning system is a finding, not a blocker.

2. Triangulate Plan and Implementation

For every claim that could change the headline status or next action:

  1. Test it against at least one independent source: the working-tree diff, commit history, actual implementation and wiring, relevant tests, generated artifacts, or recorded validation and CI results.
  2. Open the implementation entry points and tests needed to determine whether described work exists, is connected, and has validation coverage. Limit this inspection to establishing delivery state.
  3. Run a targeted check only when it is an established, reasonably fast diagnostic whose scope and side effects are understood. Prefer existing CI evidence when a local check would be slow or environment-dependent. Name important checks that remain unrun.
  4. Classify the claim:
    • Verified — independent evidence supports it.
    • Drift — present repository evidence materially contradicts it.
    • Unverified — affordable evidence is insufficient; lack of evidence alone is not contradiction.

When sources conflict, describe the document's intent separately from the observed implementation state. Current repository evidence governs the status briefing; documents may still govern desired direction.

This phase is complete when every headline claim is verified, identified as drift, or explicitly left unverified, and neither current work nor the next action rests only on a planning document.

3. Add Relevant GitHub State

When the repository has a GitHub remote and access is available:

  1. Use gh to list open pull requests with their draft state, branches, review/check state, and recent activity. Inspect the current-branch or current-work pull request more deeply when one exists.
  2. Inspect issues only when relevance is established by the current plans, commits, pull requests, an explicitly active milestone, or the user's named scope. Keep the general issue backlog outside the briefing.
  3. Connect remote work to local and documented state instead of reporting it as a separate inventory.

If the remote, authentication, or network is unavailable, finish the local briefing and state the resulting blind spot.

This phase is complete when relevant open pull-request state is included or its absence is established, and any included issue has a concrete connection to the work being reported.

4. Choose the Next Action

Recommend the smallest concrete action that best advances the observed project state. Weigh, in order:

  1. explicit user direction;
  2. coherent in-progress local or pull-request work;
  3. blockers, failing checks, or review gates preventing that work from landing;
  4. verified current planning intent;
  5. the best-supported queued plan or relevant issue.

Treat dirty or remote work as evidence of activity, not proof of ownership or priority. When two materially different actions remain equally plausible, state the decision needed instead of inventing certainty.

The recommendation is complete when it names the target, explains why it comes next, and gives a concrete first step supported by the briefing's evidence.

5. Deliver the Briefing

Lead with the bottom line: what state the project is in and what to do next. Then include only useful sections from:

  • Current state — active work, branch/worktree condition, and project health signals.
  • Recent movement — meaningful local commits, uncommitted work, and merged or active remote work.
  • Evidence and drift — material verified, contradicted, or unverified planning claims.
  • Pull requests and issues — only work relevant to the current state.
  • Blockers and blind spots — failures, missing access, ambiguity, and validation not run.
  • Recommended next action — one action and its evidence-backed rationale.

Distinguish observation, document claim, and inference. Attach concise file, commit, check, PR, or issue references to the claims they support. Avoid raw command output, exhaustive file inventories, a full issue backlog, or a general code review.

Before responding, compare repository status with the baseline and confirm that the run introduced no repository or remote changes. Remove only incidental artifacts created by diagnostics run during this invocation, and report any change that cannot be safely reversed.