agentsclimarketplace

Reflect

Skill clairem0/technically-curious-skills/self-learning/reflect

Analyze diary entries to assess how well CLAUDE.md is guiding behavior, surface new learnings worth codifying, and identify tool efficiency patterns. Run after 3-5 diary entries accumulate. Args — "last N entries", "from YYYY-MM-DD to YYYY-MM-DD", or "related to [keyword]".From its SKILL.md

Install
npx -y skills add clairem0/technically-curious-skills --skill reflect

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.6 KB, 747 tokens by cl100k_base, as published. Nobody here has run it

Reflect

Read your diary entries and ask three questions:

  1. Is CLAUDE.md still guiding well? — Review existing rules against what actually happened. Are any rules being violated? Are any rules stale, redundant, or too verbose? CLAUDE.md should stay simple and elegant — propose removals and tightening, not just additions.

  2. What new patterns would genuinely change behavior? — Look for recurring themes across entries that aren't captured in CLAUDE.md yet. The bar is high: only propose additions that would make Claude act differently in future sessions. "Nice to note" doesn't clear the bar.

  3. What tool efficiency patterns should we optimize? — These don't go in CLAUDE.md. They go into skill files as guidance or examples. Look for repeated reads, sequential calls that could be parallel, tool choice mismatches (e.g., Bash for file reads instead of Read tool).

Connectors

No external connectors needed. Uses Claude Code built-in tools (Read, Write, Glob, Grep).

Setup

Reflect reads diary entries and your CLAUDE.md, then writes a reflection document.

  • Diary entries — path from env.md (default: ~/diary/)
  • Reflections — path from env.md (default: ~/reflections/)
  • CLAUDE.md — typically ~/.claude/CLAUDE.md or your project's .claude/CLAUDE.md

Set up the diary skill first. You need at least 3 entries for meaningful pattern detection.

How This Works

Read the unprocessed diary entries (check [reflections-path]/processed.log for what's already been analyzed). Read the current CLAUDE.md. Then think across all the entries together.

For the CLAUDE.md health check: read every rule and ask — did the diary entries show this rule working? Being ignored? Being unnecessary? A rule that never comes up might be dead weight. A rule that keeps getting violated needs to be reworded or strengthened.

For new learnings: look for patterns that appear 3+ times across entries. One-off observations aren't patterns. Before proposing a new rule, ask: is this already covered by an existing rule? Could an existing rule be reworded to capture this? Addition is the last resort — tightening existing rules is better.

For tool efficiency: aggregate the tool efficiency sections across diary entries. Spot recurring inefficiencies (the same file read 5 times across sessions, sequential patterns that keep happening). These become concrete guidance in skill files — "always read config.md once at the start" or "run these searches in parallel."

Output

Write a reflection document to [reflections-path]/YYYY-MM-reflection-N.md, then present your findings.

CLAUDE.md proposals require explicit confirmation before applying. Present proposed changes clearly — removals, rewording, and additions — with reasoning. Keep CLAUDE.md simple. If you can't explain why a rule matters in one sentence, it probably doesn't.

After the reflection is complete, update processed.log so these entries aren't re-analyzed next time.

Customization Notes

  • Strength thresholds — Adjust if you generate diary entries more or less frequently. The default (3+ for a pattern) assumes weekly entries.
  • CLAUDE.md philosophy — The goal is a concise, high-signal document. Reflect should propose net-zero or net-negative changes to CLAUDE.md length whenever possible. If you're adding a rule, look for one to remove.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,679. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.