dailyaiagents-cpu

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Watcher — surfaces crons in error state. RED if >33% failing, YELLOW otherwise. Future: glob-match per-skill expectedoutputglob frontmatter for richer health.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

The convention + fault tree for proactive cross-agent help. Specialists that hit a blocked state write a delegation request to data/delegations/pending/ - .json. The tools/delegationtick.py daemon polls every 60s and dispatches to the named target agent via openclaw agent. Retry-cap 2; failures move to…

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Renders a Cooper-edited founder letter to MP3 via ElevenLabs TTS. Voice-gate + scrubber HARD before any cloud call. Pinned voice Sarah (EXAVITQu4vr4xnSDxMaL), model elevenmultilingualv2. Tracks usage against the 10K-char free-tier monthly cap.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Probe Hermes/OpenClaw/Paperclip/Obsidian. RED on outage. Auto-bootstrap if trust-graph allows (skill=gateway-pulse, domain=servicerestart, score≥0.8). Pure bash + curl + nc.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Watcher — count uncommitted files in repo, surface dominant drift class. YELLOW >100, RED >250.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Detects when workspace/systemmemory/heartbeats.json mtime exceeds the silence threshold (default 6h). Voice-alerts + Telegram-pings Cooper at high priority. Counter-measure for the v11.0 archive-too-aggressive failure that left the heartbeat-collector silent for 67h before cont-14 revived it.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Orchestrator — every 15 min, aggregate new replies across LinkedIn, Reddit, email into data/hermesinbox/ / .json. Hermes-daily-audit reads this for the funnel section. Sub-skills: linkedin-reply-poll, reddit-reply-poll, email-reply-poll.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Watcher — kickstart -k drifted critical plists. Trust-graph-gated, throttled to 1 kickstart per plist per hour.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

The skill that creates skills. Reads a problem description, asks local Ollama to extract a skill spec (name, owner, pipeline, trigger, status codes), generates scaffolded SOUL.md + SKILL.md + run.sh, runs voice-gate + scrubber + bash-syntax + smoke on a synthetic input, lands the result in /.openclaw/skills/ / if all…

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

One-shot skill that splits docs/methodology.md into 8 standalone essays, adds a 2-paragraph intro and "what to read next" footer to each, runs voice-gate (essay surface, threshold 70 HARD) plus scrubber HARD, and auto-publishes passing essays to site/essays/. Drafts that fail either gate land in data/essays/drafts/…

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Ping primary local LLM models every 4 min during business hours so cold eviction doesn't bite the next dispatcher fork. Logs to /.openclaw/logs/warmth-keeper.log. Telegrams PROBLEM only after 15+ min straight unreachable.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Aggregates monthly cash burn across Anthropic, ElevenLabs, Cloudflare, Stripe fees, GitHub Actions, and domain registrations. Reads from MCP servers where available; falls back to manual env-var or file-based entries. Output to data/finance/burn/ .md as the canonical monthly burn statement. Manual-entry rows are…

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Nightly 03:00 CT scan: identifies skills whose newest file mtime crossed 25 days. Writes stale-list to data/curator/stale-skills- .md. NEVER auto-bumps timestamps — only Cooper decides what to refresh. Sunday 09:30 CT weekly Telegram digest summarizes stale skills sorted by qualityscore (high-value first).

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Daily Stripe MRR query against a configured floor. Writes snapshot JSON to data/finance/mrrhistory/, fires Telegram high-priority alert on floor breach. Floor is 0 by default (silent) until Cooper tunes it after 30 days of data.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Daily paced outreach across LinkedIn (5 view-first DMs), Reddit (3 substantive comments + 1 DM), email (5 personalized cold emails) — 14 sends max/day. Anti-template diff ≤0.6 vs last 10 same-surface. Every send through Cooper YES/NO approval-gate. Cron M-F 10:00 CT.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Every 6h, compare loaded launchd labels (launchctl list) against source-of-truth plists in config/launchd/. Telegram-alerts on missing-from-source or missing-from-loaded. State file for cooldown so we don't repeat alerts on stable drift.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

HARD gate for public output. Scans proposed text for financial-data leaks (dollar amounts, MRR/revenue/P&L terms, customer counts, kalshi.com/profile URLs). Allows known-good config values (Path E parameters, sticker prices) via context-sensitive allowlist. Blocks publish on ANY hit. Cooper override required to…

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Hourly publish of an operational ticker — last 24h commits, voice-gate pass/fail counts, P-gate catches, smoke-tests, substrate state. Auto-publishes to site/receipts.html + site/api/receipts.json. Public surface, scrubber-HARD. Operational receipts compound into a recruitable artifact for prospects who want to verify…

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Daily 09:00 CT browser-session audit. Probes 3 default targets (linkedin, linkedin, reddit) via Chrome CDP at 9223. URL-redirect detection (more robust than DOM selectors). Writes /.openclaw/state/session-keeper/ .json. Opens approval gates for expired sessions with the exact CDP-spawn command Cooper taps.…

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Friday 16:00 CT auto-draft of the weekly ship log. Aggregates 7 days of git commits + auto-promoted skills + postmortems + improvement-queue entries into data/ship-log/ -draft.md. Voice-gates (advisory, score 70+) and scrubber-checks (HARD gate). Telegrams Cooper to fill the "What I learned" paragraph before running…

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Walks every skill directory in skills/ and /.openclaw/skills/, validates each has SOUL.md (voice ≥75), SKILL.md, run.sh (executable), and a launchd plist OR an "on-call" marker. Outputs a coverage report listing missing pieces. Run weekly + on-demand.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Append-only telemetry helpers — trackstart and trackend. Adds two-line instrumentation to any skill so the catalog produces measurable invocation counts, success rates, and median wallclock. Read by skill-stats reporter daily to identify dead skills (0 invocations in 30d) and hot skills (>10/day).

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Auto-wire newly-shipped skills into owner-agent SOUL.md by reading owneragent: frontmatter and rewriting the block between AUTO-GENERATED-SKILLS markers. Eliminates the hand-edit-or-silently-fail pattern. Cron 0 /6 CT.

dailyaiagents-cpu/dailyai-os

Skill Claude CodeCodex

Daily 09:00 CDT. Picks a skill from the catalog (weighted random; prefers skills not spotlit in last 60 days; prefers skills with measurable outcomes). Reads the SOUL.md plus SKILL.md and asks local Ollama qwen3.5 to write a 200-word non-technical explainer for a solo founder who hasn't seen DAI OS. Voice-gate…