Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/xcota/pos/reflectnpx skills add xcota/pos --skill reflectgit clone --depth 1 https://github.com/xcota/posWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00019 | $0.01103 |
| Opus 5 | $0.00010 | $0.00551 |
| Sonnet 5 | $0.00004 | $0.00221 |
| Haiku 4.5 | $0.00002 | $0.00110 |
Grade A, and why
reflect scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/reflect — Periodic Self-Reflection
Analyze recent work, extract patterns, learn from mistakes, propose improvements.
Core principle: the strongest self-improvement signal is NOT the daily summary — it is the conversation itself: what the user says, in what tone, where friction or correction shows up; and how the agent responds (where it landed, where it substituted its own agenda, failed to orient, or over-produced). Mining recent transcripts for that friction is the crown jewel of this skill.
When to Use
- Manually: when you want to review and improve
- Auto: wire up a periodic trigger yourself (e.g. a cron job or
/loopcalling/reflect) — not shipped by default
Steps
- Gather data:
- Read last 3-5 daily notes from
daily/ - Read recent entries in
state/decisions/ - Read
context/anti-patterns.md(what were we watching for?) +anti-patterns-index.mdwatch-list - Read
context/learned.md(what did we already know?) - Check recent git log:
git log --oneline -20
- Read last 3-5 daily notes from
1b. Analyze the CONVERSATION (the main signal):
- Source = recent session transcripts. Resolve the path generically from this project's slug under the Claude projects dir, e.g.:
(Adjust the slug derivation to your harness; the point is workspace-derived, not hardcoded.) Do NOT bulk-load — extract only user turns + agent replies + friction points.PROJECT_SLUG=$(pwd | sed 's#/#-#g') TRANSCRIPT_DIR="$HOME/.claude/projects/$PROJECT_SLUG" NEWEST=$(ls -t "$TRANSCRIPT_DIR"/*.jsonl 2>/dev/null | head -1) - Extract user turns, e.g.:
(Adapt to the actual transcript format.) Look for: corrections, frustration/anger (friction), "not that / not like that", energy shifts, what the user re-asks.cat "$NEWEST" | python3 -c "import json,sys; [print('U:', json.loads(l).get('message',{}).get('content','')[:300]) for l in sys.stdin if json.loads(l).get('type')=='user']" - About the USER: how they respond (tone, length, what they value/dislike in THIS session), what latent needs surfaced → candidates for
context/identity.md/ memory, with a source-tier. - About the AGENT: where it landed, where it substituted its own agenda (substitution-instinct), where it failed to orient before committing, over-produced or added noise, made the user wait. Each real friction point → anti-pattern candidate.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 88 lines · 19 tokens per session scan A be0bd957581a
reflect is a skill published in the GitHub repository xcota/pos (43 stars, last pushed 1mo ago), licensed MIT. It adds 19 tokens to every session and 1,103 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
second-brain-climb
为一个长期方向维护深入、由证据支撑的‘爬山’模型，包括山顶、当前位置、差距、路线、约束、阻点、证据、不确定性、下一信息行动、当前押注、决策标准、检查点、反思和真实工作链接。用户讨论长期追求、人生方向、山、山顶、战略、路线选择、进展停滞、继续还是转向，或要求更新个人 AI 基底中的山脉时使用。.
ori-memory
Persistent agent memory with learning retrieval. Knowledge graph on markdown files — capture insights, decisions, research, and learnings during work, then retrieve them weeks or months later. Use when knowledge is too valuable to lose but too much to inject into every prompt.
sibyl
Persistent memory and task coordination for this project. Invoke for any prompt about past work, project state, in-progress tasks, prior decisions, gotchas, or capturing a non-obvious learning. Also covers semantic search across project knowledge and external docs.
day-log
Quick timestamped log entry — no conversation, just capture.
period-summary
Period summary (w/m/q/y) — reflect on the period, score it, close it out.
day-start
Day briefing — deliver context, capture what's happened, get moving.