Skills de Claude Code · página 72

Post-batch audit — checks whether all enabled scheduled skills fired in their expected window, alerts on silent misses, files issues on batch-level outages

Multi-beat news thread tracker — persists beat counts per active storyline, searches for new developments, alerts when a thread hits the article-ready threshold (3rd beat)

Map of who's building on top of watched repos — forks, third-party ecosystem repos, builder announcements. Answers the "who's building on top" question.

Read-only audit of installed skills' capability coverage — maps every enabled/disabled skill against the locked 6-value taxonomy in docs/CAPABILITIES.md, flags any capability tier with zero enabled coverage as an actionable gap

One-shot Phase 2 sweep — infers a `capabilities:` declaration for every skill that doesn't have one yet by pattern-matching its SKILL.md body against the locked taxonomy, writes a JSON proposal manifest, and opens a single PR adding the declarations. Reruns safely (skills already declared are skipped). Closes the ~174-skill undeclared backlog left after the Phase 1 high-blast-radius sweep (PR #322).

Recap article from a public Telegram channel — rank posts by engagement, expand on the best

Report on TODOs, dead code, and test coverage gaps

Scan of Product Hunt + Hacker News for NEW AI-agent-framework launches outside the 9-framework cohort already tracked by ai-framework-watch

Tracker for the AI compute market — GPU/hardware deals, inference pricing trends, decentralized compute token signals, and lab vs hyperscaler dynamics.

Validate aeon.yml and .github/workflows/aeon.yml for structural invariants that have caused past outages — checkout step ordering, duplicate skill keys, missing skill files

Operator tweet performance tracker — engagement metrics, top formats, topic resonance; closes the content feedback loop in the article/tweet production pipeline

Audit any contract on Base — verification, proxy/upgradeability, ownership/admin roles, and mint/freeze/pause/drain powers as a live capability matrix. Keyless via Etherscan v2 + Base RPC.

Closes the contributor flywheel — turns the contributor-leaderboard ranking into a tier-priced rewards plan, writes it into memory/distributions.yml, and hands off to distribute-tokens for the actual on-chain send

Recognition post for one fork operator — converts fork-cohort cohort data into a named human moment (POWER fork callout with their work, stars, and skills enabled)

API cost report — computes dollar costs from token usage, flags anomalies, forecasts burn, and prescribes concrete optimizations

Provision Meta campaigns and ad sets on AdManage.ai from a declarative config. Runs on-demand — creates entities PAUSED, writes the returned IDs back into state so schedule-ads can launch into them.

Generate a complete new skill from a one-line prompt and ship it as a PR

Funding round tracker across configurable verticals

Exhaustive multi-source synthesis on any topic with explicit source credibility tiering and per-finding confidence — analyst-grade, not aggregator-grade

Check pool health, positions, and yield rates for tracked protocols

DeFi read — regime verdict, biggest movers with "why it matters", sustainable vs incentive yields, fees fundamentals

Generate a small app or tool and deploy it live to Vercel via API

Map every contract deployed by an address on Base, link reused patterns, and surface serial-rug signals. Keyless via Etherscan v2 + Base RPC.

Generate and send a digest on a configurable topic

Audit of the pending vulnerability disclosure queue — tracks draft advisories in memory/pending-disclosures/, alerts on aging CRITICAL/HIGH findings.

Send tokens to a list of contributors via Bankr Wallet API with per-recipient idempotency, two-phase resolve→execute, dry-run, and recovery from partial runs

Diff of ECOSYSTEM.md against the last run — surfaces newly-added projects (and projects removed) as a discrete signal so new entrants aren't lost in the static list. Pairs with ecosystem-pulse (liveness).

Link-health audit of ECOSYSTEM.md — checks every GitHub repo for archived/disabled state and every project URL for HTTP 4xx/5xx or redirect chains, surfacing dead/archived/moved entries before a casual reader stumbles into one. Closes the three-skill ecosystem loop with ecosystem-entrants (arrivals) and ecosystem-pulse (liveness).

Liveness check of the projects listed in ECOSYSTEM.md — stars/forks/last-commit recency + new releases for any project that can be matched to a GitHub repo

Turn flagged engagement opportunities into ready-to-post replies — read recent logs, draft specific responses, send as copy-paste-ready output

Proactively enhance watched repos — fix issues, add features, improve code

Clustered, signal-scored digest of Farcaster casts with conversation-shape lead and insight-first editorial notes

Conditional scan — fires only when Fear & Greed < 25. Identifies assets outperforming during broad market fear, synthesizes narrative catalysts from memory, and delivers a terse conviction setup brief. Skips silently when market conditions don't qualify.

Build a feature for every watched repo in one run — iterates the full repo list, picks one feature per repo from yesterday's repo-actions ideas first

Search X/Twitter for tweets about a token, keyword, username, or topic — clustered by sub-narrative

Monitor managed Aeon instances — check health, dispatch skills, aggregate status

Fleet-wide scorecard across this instance and every managed instance in memory/instances.json — runs, tokens (OpenRouter shape), est. cost, skills, and reliability, with deltas vs the previous run and alerts

Fleet skill-adoption leaderboard — per-slug count of how many POWER+ACTIVE forks have each upstream skill enabled, top-15 most-adopted and bottom-15 least-adopted by fleet penetration, silent when nothing moves

Fleet-state digest — synthesises fork-cohort, contributor-spotlight, and fork-release into one "state of the fleet" narrative

Escalation audit — parses the follow-up / open-loop section of MEMORY.md plus the issue tracker, computes item ages, and alerts on items hitting urgency thresholds so nothing rots unattended

Fork-activation cohort tracker — buckets every fork by recent run activity (COLD / STALE / ACTIVE / POWER) with deltas since the prior snapshot

Ranking of developers contributing to the fork fleet and back upstream

Named alert when a fork completes its first ever workflow run — catches the activation moment that slower fork-cohort snapshots would miss

Inventory active Aeon forks, detect diverged work, surface upstream contribution candidates

Per-fork health tier synthesizing push recency + enabled skill count + 30d PR activity into ACTIVE/WARM/STALE/QUIET buckets; fleet health ratio + top-10 ACTIVE table; silent when nothing moves

Scan that celebrates when any fork of the parent repo cuts a tagged GitHub release. Silent when no fork releases in the window.

Cross-fork customization digest — surfaces where the fleet's enable/disable/var/model decisions diverge from upstream defaults

Cross-fork skill adoption digest — per-fork table of upstream skills the fork hasn't enabled, top forks by gap size called out, silent when gaps are small

Trace where funds move to (or came from) across multiple hops from a Base address and render a Mermaid flow graph. Keyless — no explorer key needed.

Digest of new open issues across your repos, ranked by signal (security/bug/feature/other)

Watch repos for stale PRs, new issues, and new releases — tiered by urgency with concrete next actions

Upgrade-triage digest of new releases across watched AI/infra/crypto repos

Curated trending GitHub repos — clustered, filtered, and labeled by momentum

Compare current progress against goals with quantified status, velocity, trend, and a concrete next action per goal

Top HN stories filtered by interests, with comment-mined insights and themed clustering

Proactive ambient check — surface anything worth attention

Analyze token holder distribution on Base — top-N share, HHI concentration, LP/lock/burn exclusions, and whale clusters. Keyless via Etherscan v2 + Base RPC.

Detect un-sellable / restricted (honeypot) tokens on Base by simulating a real holder's sell via eth_call. Keyless — no explorer key needed.

Curated trending Hugging Face models, datasets, and spaces — filtered, clustered, and labeled with a "why notable" line per pick

Quick note capture triggered via Telegram — restates, triages, logs, and echoes back for confirmation

Pre-batch API provider health check — detects credit exhaustion or auth failure for every configured provider key before the scheduled batch runs, giving the operator a window to act before skills degrade

Diff of the Atrium marketplace catalog at https://atriumhermes.tech/.well-known/skills/index.json against the prior snapshot — surfaces newly-published skills, removed skills, and updated descriptions. Supply-side complement to sparkleware-catalog (curated skill-packs.json registry) and skill-update (version drift of installed skills).

Scan of Product Hunt + Hacker News for NEW AI-agent-framework launches outside the 9-framework cohort already tracked by framework-watch

Ranking of developers contributing to the fork fleet and back upstream

Conditional scan — fires only when Fear & Greed < 25. Identifies assets outperforming during broad market fear, synthesizes narrative catalysts from memory, and delivers a terse conviction setup brief. Skips silently when market conditions don't qualify.

Escalation audit — parses the follow-up / open-loop section of MEMORY.md plus the issue tracker, computes item ages, and alerts on items hitting urgency thresholds so nothing rots unattended

Cross-fork customization digest — surfaces where the fleet's enable/disable/var/model decisions diverge from upstream defaults

Named alert when a fork completes its first ever workflow run — catches the activation moment that slower fork-cohort snapshots would miss

Per-fork health tier synthesizing push recency + enabled skill count + 30d PR activity into ACTIVE/WARM/STALE/QUIET buckets; fleet health ratio + top-10 ACTIVE table; silent when nothing moves

Scan that celebrates when any fork of the parent repo cuts a tagged GitHub release. Silent when no fork releases in the window.

Competitive-intelligence digest on the AI agent framework space — momentum, releases, breaking changes across a curated watchlist

Per-skill run-count watchdog — checks if any capped skills exceeded their configured daily limit and alerts on breach

Top HN stories filtered by interests, with comment-mined insights and themed clustering

Execution-gap audit — cross-references the startup idea backlog against shipped skills, prototypes, and cross-repo PRs. Surfaces the top 3 ideas to build next based on narrative fit and operator fit.

Route a prompt through a BEAMR x402 gateway and report the answer + onchain receipt

Build on-chain automation workflows on Base via CTRL. Use for recurring or triggered actions — DCA, price-gated swaps, launchpad sniping, whale-follow — that should run autonomously after a single wallet signature. The wallet signs once (EIP-5792 batch), and the CTRL keeper executes every trigger after, bounded by per-swap and per-day caps the user pre-authorized.

>-

>-

>-

>-

>-

>-

>-

>-

>-

>-

Use when retiring old logic, collapsing duplicate owners, removing fallbacks, or touching schema, persistence, or source-of-truth boundaries while deciding whether to delete old paths, retain compatibility, or stop for confirmation.

Use when defining new features, product behavior, UI/component design, architecture choices, contract changes, or ambiguous medium/high-complexity work before implementation.

Use when the user asks for caveman mode, fewer tokens, brief responses, compressed communication, or otherwise explicitly requests a much shorter answer.

Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies

Use when entering a project for the first time, or when the user asks to establish shared language, define domain terms, or create a project glossary.

Use when you have a written implementation plan to execute in a separate session with review checkpoints

Use when implementation is complete, verification has passed, and the user needs to choose merge, PR, branch cleanup, or follow-up integration handling.

Use when the user explicitly asks for first principles, first-principles review, Occam's razor, or when a complex decision has ambiguous goals, competing constraints, repeated fixes, fallback growth, duplicate owners, or architecture/product direction risk.

Use when the user explicitly sets an Aegis goal with /aegis-goal, Aegis goal:, or asks to define goal, success evidence, stop condition, or task boundaries before work.

Use when a task is multi-step, may span context resets or sessions, uses subagents, or risks losing state before completion.

Use when receiving code review feedback before implementing suggestions, especially when feedback is unclear, risky, disputed, or technically questionable.

Use when the user asks to create, write, update, amend, supersede, or evaluate an ADR, architecture decision record, durable architecture decision, decision log, or baseline sync after architecture-changing work.

Use when explicitly requesting an independent code review, after subagent-driven implementation slices, before merging high-risk work, or when verification finds evidence, baseline, architecture, compatibility, or retirement uncertainty that needs reviewer scrutiny.

Use when executing implementation plans with independent tasks in the current session