Add retrospective orb sizing skill

[maphew]

Summary

• add an Amp skill for evidence-based retrospective orb sizing
• analyze exported thread shell results for real resource pressure and GNU time measurements
• avoid false recommendations when telemetry is absent or transcript text merely discusses OOMs
• bootstrap fresh-orb Beads state from authoritative refs/dolt/data before JSONL auto-import can occur
• keep setup idempotent across normal clones and linked Git worktrees

Verification

• python3 -m unittest discover -s .agents/skills/analyzing-orb-sizing/tests -v
• python3 -m py_compile .agents/skills/analyzing-orb-sizing/scripts/analyze.py
• bash -n .agents/setup
• ran .agents/setup twice successfully after correcting its linked-worktree guard
• recovered authoritative Dolt state and completed a normal bd dolt push

Amp thread: https://ampcode.com/threads/T-019f8cb6-a6f7-77ab-a6d7-ea71e143dc9d

amp-gpt-5.6-medium on behalf of maphew

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: da28ede66c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[maphew]

Pushed 91a3d26 addressing the review sweep findings:

1. The analyzer now accepts Amp's current Bash tool calls (cmd) and top-level tool_result.output / status, including structured text output, while retaining compatibility with legacy shell_command and nested run.result exports.
2. Hard pressure at a0.large now explicitly says that it is the largest documented size and directs workload investigation or a larger-orb request instead of silently implying that the same size is a fix.
3. Setup validates the exact $HOME/.local/bin/bd binary it later invokes, so an unrelated pinned bd elsewhere on PATH cannot cause bootstrap to fail.

Regression coverage increased from 8 to 11 analyzer tests. Validation passed: analyzer unit tests, Python compilation, bash -n, git diff --check, cargo fmt --check, clippy with warnings denied, and the full Rust test suite (63 tests).

codex-gpt-5.6-sol-medium on behalf of maphew

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (3 files)

• .agents/setup
• .agents/skills/analyzing-orb-sizing/scripts/analyze.py
• .agents/skills/analyzing-orb-sizing/tests/test_analyze.py

Previous Review Summary (commit 91a3d26)

Current summary above is authoritative. Previous snapshots are kept for context only.

Previous review (commit 91a3d26)

Status: 3 Issues Found | Recommendation: Address before merge

Overview

Severity	Count
CRITICAL	0
WARNING	3
SUGGESTION	0

Issue Details (click to expand)

WARNING

File	Line	Issue
.agents/skills/analyzing-orb-sizing/scripts/analyze.py	199	cpu_matches[-1] picks next report's CPU value when multiple GNU time reports share one output; use [0]
.agents/skills/analyzing-orb-sizing/scripts/analyze.py	147	Current-schema exit codes collapsed to 0/1/None; direct exit_code == 137 check never fires for current-schema exports
.agents/setup	32	Guard checks main repo's .beads but bootstrap targets $repo_root (worktree); inconsistent for linked worktrees

Files Reviewed (4 files)

• .agents/setup - 1 issue
• .agents/skills/analyzing-orb-sizing/SKILL.md - 0 issues
• .agents/skills/analyzing-orb-sizing/scripts/analyze.py - 2 issues
• .agents/skills/analyzing-orb-sizing/tests/test_analyze.py - 0 issues

Fix these issues in Kilo Cloud

---

_{Reviewed by efficient · Input: 60.2K · Output: 16.3K · Cached: 129.2K}

[maphew]

Pushed 61bf5f6 after the current-head Kilo review.

• Fixed multi-report GNU time parsing so every RSS value is paired only with the CPU and elapsed fields that precede that same report; added a two-report regression test with deliberately different values.
• Fixed linked-worktree setup so the guard and bd bootstrap both target the common repository root.
• Investigated the exit-137 warning. Amp's documented current ThreadToolResultBlock schema provides status and output, but no numeric exit code. The analyzer already recognizes OOM, signal 9, and textual exit-137 evidence from failed output. Treating every generic error as resource pressure would create false positives, so no numeric status is invented when the export does not carry one; exact numeric handling remains available for legacy exports that do carry exitCode.

All 12 analyzer tests, Python compilation, shell syntax, and git diff --check pass.

codex-gpt-5.6-sol-medium on behalf of maphew