dayflow-pull
Pull Henry's Dayflow activity data (what he actually worked on, day by day, card by card, with accurate hours) straight from Dayflow's local SQLite DB — read-only. Use whenever Henry asks to check his Dayflow, compute real work hours, build a weekly/daily report of what he did, verify time spent on a project, or audit his activity. Computes overlap-merged wall-clock hours (Dayflow's own sums double-count) and flags corrupt cards instead of trusting the LLM summaries at face value.
- License 许可
- MIT
- 所属仓库
- caezium/skills
- 最近更新
- 2026-08-24
- Stars
- ★ 3
- Forks
- 0
01 前言 Front Matter
- name
- dayflow-pull
- description
- Pull Henry's Dayflow activity data (what he actually worked on, day by day, card by card, with accurate hours) straight from Dayflow's local SQLite DB — read-only. Use whenever Henry asks to check his Dayflow, compute real work hours, build a weekly/daily report of what he did, verify time spent on a project, or audit his activity. Computes overlap-merged wall-clock hours (Dayflow's own sums double-count) and flags corrupt cards instead of trusting the LLM summaries at face value.
02 资源文件 1
- 脚本pull.py5.2 KB
03 SKILL.md 正文
dayflow-pull
Read Henry's Dayflow data directly from its structured store — do not eyeball the app or trust Dayflow's own hour totals. Dayflow is a screen-recording time tracker whose cards are LLM/OCR-generated, so two things must be handled:
- Overlapping cards — Dayflow sometimes emits duplicate/overlapping cards
for the same span (esp. after re-analysis).
SUM(end_ts - start_ts)double-counts them. Always interval-merge for real wall-clock hours. - Corrupt timestamps — occasional cards span days (a "System" card of 1400+ hours was observed). Exclude any single card longer than ~3h and flag it.
How to use
Run the helper (read-only; safe while Dayflow is running):
python3 ~/.claude/skills/dayflow-pull/pull.py # last 8 days, Work, per-day merged hours + cards
python3 ~/.claude/skills/dayflow-pull/pull.py --from 2026-07-08 --to 2026-07-14
python3 ~/.claude/skills/dayflow-pull/pull.py --day 2026-07-13 # one day, card by card
python3 ~/.claude/skills/dayflow-pull/pull.py --category all # every category, not just Work
python3 ~/.claude/skills/dayflow-pull/pull.py --obs # also dump the rawer `observations` layer
Categories: Work (default), Personal, Distraction, Idle, System, all.
Data model (for ad-hoc queries)
DB: ~/Library/Application Support/Dayflow/chunks.sqlite — open mode=ro only.
timeline_cards— the summary cards:day,start_ts/end_ts(unix),title,summary,detailed_summary,category,subcategory,metadata(JSON incl.appSites),is_deleted(filter=0).observations— the finer, rawer layer beneath the cards (per-batch), joined viabatch_id; use to sanity-check what a card claims vs. the underlying evidence (llm_modelrecords which model produced it).llm_calls— Dayflow's own LLM invocations, if you need to audit how a card was derived (the "OCR-LLM might be wrong" check).journal_entries,day_goals,daily_standup_entries— journal/goals.
Rules
- Read-only, always.
sqlite3 -readonly/?mode=ro. Never write; the DB is live. - Report merged hours, not summed. State when you excluded glitch cards.
- Read the cards, don't hand-wave. Cite actual card titles/times, and drop to
observationswhen a card's claim looks off — that's the point of reading the structured data instead of guessing from a summary.
04 同仓库更多 Skills
writing-for-agents
Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.
wrangler
Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.
workers-best-practices
Reviews and authors Cloudflare Workers code against production best practices. Load when writing new Workers, reviewing Worker code, configuring wrangler.jsonc, or checking for common Workers anti-patterns (streaming, floating promises, global state, secrets, bindings, observability). Biases towards retrieval from Cloudflare docs over pre-trained knowledge.
wizard
Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.
why
Use for 'why does X work this way', 'why we picked Y', design rationale, regressions, postmortems, or data-backed thresholds. Discovers available evidence sources, searches each relevant category, and returns a cited read on decisions and tradeoffs. Use how for runtime behavior.
week-prep
Generate a briefing with calendar, tasks, and attention flags. Use for "week prep", "briefing", or "what's on this week".