AI Agent Workflow Collection

DB + filesystem contract every Scrittore follows when taking a position from scored (≥50) to ready/excluded. Three gates BEFORE writing a single line of CV (anti-rewriting, anti-collision, link verification), one canonical path for deliverables, one final gate after the 3rd Critic round. Skipping any of these produces duplicate work, overwrites another Writer's claim, or — worst — pushes an excluded-grade CV to the user as ready. Owned by the Scrittore.

Blind peer review for code - multiple agents critique independently, then synthesize privately to avoid…

Drain pending bridge verdicts at the START of every Captain turn — MANDATORY first action before doing anything else. During a long turn, jht-tmux-send from the bridge can fail with rc=3 (text never appeared in the pane) and a [BRIDGE PACING] or PIPELINE STALLED verdict gets silently dropped. The bridge appends EVERY verdict to a JSONL mailbox so you can recover them. Skipping this drain means acting on stale measurements while a fresher verdict sits unread.

Translate a [BRIDGE PACING] 15-min calibration tick into per-agent throttle adjustments. The bridge measures the team's actual rate of consumption and gives you a verdict (SFORO / MARGINE / ALLINEATO) plus the per-agent share + cadence needed to choose WHO to slow down and BY HOW MUCH. Open this skill ONLY when a [BRIDGE PACING] line lands; the routine [SENTINELLA] orders use a different flow (sentinel-orders).

Reclaim disk on the shared JHT caches (uv wheel cache + codex SQLite log) every ~24h. Owned by the Dottore — single-instance, runs at the end of a routine round when the team is idle. Never run mid-emergency: the SQLite VACUUM blocks for ~30s on a 200 MB DB and would steal cycles from a Sentinel-driven recovery. Migrated from the Captain so the Captain stays focused on coordination, not housekeeping.

Daily handoff diary for the Captain. The Captain is restarted often (context-refresh, new work window, reboot) and otherwise loses the day's hard-won pacing lessons — repeating the same mistakes (e.g. 3 Scouts at once → an unbrakable spike → a 5h coast to repay the debt). Read the PREVIOUS day's notes at startup (handoff), and APPEND a one-line note whenever something significant happens during the day (a scaling decision, a spike, a kill, a lesson). One append-only file per day.

Check provider usage via HTTP fast path (~2s, zero CLI tokens). PRIMARY action when the bridge fails to read usage. Falls back to check-usage-tui if HTTP returns RATELIMIT or fails.

Strategy map for what to search WHERE, derived entirely from the candidate profile. The 5 concentric circles (workmode + relocation) tell you the geographic scope; the 4 source tiers (LinkedIn → ATS aggregators → niche → web) tell you which platforms to drain in order. A scout that searches the wrong tier in the wrong circle wastes its quota and its scout-coord partition. Open this skill at boot (after scout-coord) and again whenever a circle is exhausted or a [FEEDBACK] from the Analyst suggests changing source.

Run the mandatory 3-round CV review loop with the Critico — autonomously, without going through the Capitano. For each round you spawn a FRESH CRITICO-S session (same N as your Scrittore session: SCRITTORE-2 → CRITICO-S2), send PDF + JD, wait for the structured verdict, kill the Critic, correct the CV, regenerate the PDF, and start the next round with another fresh instance. Three rounds are non-negotiable — neither 1 nor 2. After the 3rd round, gate: criticscore ≥ 5 → ready, else excluded. Owned by the Scrittore.

Write the CV markdown that will be PDF'd and reviewed by the Critico. First read the user's ORIGINAL CV in $JHTHOME/profile/sources/ to mirror their voice and match the register to the industry (the tech defaults — "Technical Skills: Python", metric+tech-in-parens — are wrong for hospitality/sales/care/creative). Six fixed sections, max 2 pages, every claim traceable to candidateprofile.yml/sources (zero invenzioni — T10). Languages appear ONCE (Header only). Cover Letter only if the JD explicitly asks, with its own "Cover Letter" title (never "CV"). Owned by the Scrittore. Pair with application-flow (claim + path) and critic-loop (review iterations).

Pre-emptive mass-restart of every team agent once per 24h for context freshness. Owned by the Dottore. Runs only inside a narrow daily window (default 03:00 UTC ± 30 min) and only if no wave has fired in the last 23h. Each agent is killed + respawned via the same atomic sequence of liveness-check Step 3, ordered tier 3 → tier 2 → tier 1 so the workers cycle first and the coordinators (Capitano/Sentinella/Mentor/Assistente) last. Background: Codex/Kimi long-lived sessions accumulate \"noise\" — old decisions, stale facts, prompt drift — and become measurably less lucid after hours. Empirical evidence from Case Study 1 (Codex run 2026-05-19/21): manual mass-restart restored decision quality. This skill closes that gap without manual intervention.

Insert NEW records into the JHT DB (positions / scores / applications / companies / positionhighlights). Use it ONLY when an agent needs to create a record — Scout for positions, Analyst for companies and highlights, Scorer for scores, Writer for applications. Never blind-overwrite — for updates use db-update.

Reference table mapping proj (projected usage at reset) to a Sentinel state and a throttle level (0-4). Use it every tick AFTER you have a fresh sample to decide which order to send to the Captain.

How to handle rate-limit emergencies and the FATAL cascade when the bridge goes blind. Includes the cooldown-bypass triggers, the L4-SOFT/L5-HARD recovery path, and the RESET SESSIONE handling on a >30-point usage drop.

Read user feedback (like/dislike/hide/star) for a given position from the cloud. Used by the Scorer to apply a multiplier on the final score and by the Scout as a contextual signal. Returns a neutral "no signal" payload when cloud is disabled or unreachable, so callers never hard-fail.

Convert UTC timestamps to the user's timezone before showing them in chat, charts, Telegram, or any user-facing output. Use this helper whenever you would otherwise write a raw strftime("%H:%M") of a UTC datetime in something the user reads.

Offer 2-5 context-specific clickable reply buttons in the JHT game chat when they genuinely make the user's next decision easier. Use only for a small bounded choice; otherwise answer normally with jht-send. Never use these as a fixed onboarding tree.

Diagnose whether a team agent's tmux session is alive, in a long turn, or silently dead — and respawn it preserving context if dead. Owned by the Dottore (the team's roving health-check agent), not by the Captain. The core failure mode this skill catches: jht-tmux-send returns exit 0 even when the target CLI has crashed (the message is written into a bare bash, then lost). Without periodic liveness checks the team keeps \"talking to a corpse\" and the Captain counts on actions that will never happen.

Standardize positions.location free-text into structured loc/work/rolefamily columns BEFORE marking any position as checked. Covers 10 special cases (Europe Remote, Italy+remote, multi-location, US-entity-in-EU). Enforces one-position-at-a-time, peer-aligned vocabulary, never-NULL workcountry. Use whenever the Analyst is about to set status=checked on a position.

Extract the company logo for a company in the companies table and store it as a small base64 data-URI (max ~35KB, min 32px). Primary path is fully automated via logofetch.py against the official website (apple-touch-icon → icon → og:image → favicon); when the site blocks bots or has no usable icon, find a direct logo image URL via web search and pass it with --from-url. Verify the website really belongs to the company BEFORE fetching. Sets companies.logo, logosource, logofetched.

State variables the Sentinel must keep across ticks (last order sent, freeze flag, FATAL streak, per-state counters, emergency history, cooldowns). Use this skill on every tick to update memory and decide if a new order is required (edge-triggered).

How the Mentor speaks once a pattern from mentor-patterns has crossed the threshold. Three output formats — strategic advice (rare, weighty), weekly digest, on-demand answer — each with strict shape and voice rules. The Mentor's authority comes from how rarely the words come and how heavy each one weighs; this skill enforces that. Owned by the Mentor. Pair with chat-web (delivery via jht-send) and mentor-patterns (the trigger).

The five patterns the Mentor hunts in the records to decide WHEN to speak. Silence is the default; only a real, recurring pattern earns a word. This skill gives the canonical detection method for each pattern (DB query + threshold) so the Mentor never speaks from a single data point. Read-only — never writes to the DB. Owned by the Mentor.

Geocode the precise office building (lat/lon/address) for a position AFTER location-enrichment has populated loccity/loccountry. Use web search aggressively (3+ attempts) to find the company HQ/office address, then resolve coordinates via Nominatim/Photon. Skip ONLY after exhaustive search fails or when multiple ambiguous offices exist. Sets officelat, officelon, officeaddress, officegeocoded, officeverified.