Skip to content

Spec 285: Copilot Chat integration best-practices upgrade#676

Open
IanMayo wants to merge 3 commits into
mainfrom
claude/llm-debrief-next-steps-oeyg8k
Open

Spec 285: Copilot Chat integration best-practices upgrade#676
IanMayo wants to merge 3 commits into
mainfrom
claude/llm-debrief-next-steps-oeyg8k

Conversation

@IanMayo

@IanMayo IanMayo commented Jul 11, 2026

Copy link
Copy Markdown
Member

Summary

Adds the specification for feature 285-copilot-best-practices — the successor to spike #284 (Copilot Chat drives Debrief). With the local-model experiment deliberately deferred (#235), this feature upgrades the Copilot integration to current VS Code/Copilot best practice to optimise what we're demoing and researching.

Five prioritised user stories:

  1. P1 — Domain priming actually loads: the spec(113): Add flip-card interaction specification for Log Panel #284 priming file sits at apps/vscode/.github/, which the host ignores unless it's the workspace root; fix placement for the preview workspace and repo root, with a canonical source + drift check.
  2. P1 — Reproducible demos: a "Debrief Analyst" custom agent (pinned model, restricted tool list) plus the eight spec(113): Add flip-card interaction specification for Log Panel #284 scenarios as slash-command prompt files.
  3. P2 — Approvals hardening: the mutating run-tool becomes ineligible for auto-approval (bypass-proof mutation gate); documented pre-approval makes the three read-only tools dialog-free for demos.
  4. P2 — Budget-aware plot summary: compact serialization, priority-ranked budget-adaptive content replacing the fixed 200-feature cap, and a per-feature spatial digest closing the "northern track" Q2 gap from the spec(113): Add flip-card interaction specification for Log Panel #284 findings.
  5. P3 — MCP hybrid experiment: register the existing Python STAC MCP server with agent mode, compare MCP-served vs extension-served catalog search, findings feed the E13 tool-surface contract.

Informed by the #284 findings report (specs/284-copilot-plot-editing/evidence/findings-report.md) and current platform guidance (MCP GA, virtual tools / 128-tool grouping, custom agents, per-tool approval policy).

Contents

  • specs/285-copilot-best-practices/spec.md — feature specification (5 user stories, 24 FRs, 8 success criteria, assumptions, out-of-scope)
  • specs/285-copilot-best-practices/checklists/requirements.md — spec quality checklist (all items pass; no NEEDS CLARIFICATION markers)

Spec only — no code changes. Next step: /speckit.clarify or /speckit.plan.

🤖 Generated with Claude Code

https://claude.ai/code/session_01HeMQ7e9WSeBZ4aSmaHcbzQ


Generated by Claude Code

Successor to spike #284. Covers priming-file placement fix, Debrief
Analyst custom agent, prompt-file demo scenarios, budget-aware
token-efficient tool results with per-feature spatial digest,
approvals hardening, and the MCP hybrid catalog-search experiment.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HeMQ7e9WSeBZ4aSmaHcbzQ
@github-actions

github-actions Bot commented Jul 11, 2026

Copy link
Copy Markdown
Contributor

📋 Spec Navigator

Review this PR's specs interactively:
👉 Open in Spec Navigator
Raw spec directories changed in this PR:

@IanMayo IanMayo temporarily deployed to debrief-preview-pr-676 July 11, 2026 14:47 Inactive
Phase 0-2 outputs: research.md (R1-R10 decisions), plan.md with
constitution check, data-model.md, contracts (customization pack,
plot summary v2), quickstart.md, cached opening context. Agent
context updated.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HeMQ7e9WSeBZ4aSmaHcbzQ
@IanMayo IanMayo temporarily deployed to debrief-preview-pr-676 July 11, 2026 15:42 Inactive
46 tasks across 8 phases: setup + foundation (drift check), US1 priming
placement, US2 curated agent + scenario prompts, US3 approvals hardening,
US4 budget-aware summary + spatial digest, US5 MCP hybrid experiment,
polish (evidence + blog + PR).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HeMQ7e9WSeBZ4aSmaHcbzQ
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants