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/luiseiman/dotforge/watch-upstreamnpx skills add luiseiman/dotforge --skill watch-upstreamgit clone --depth 1 https://github.com/luiseiman/dotforgeWhat 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 | $0.00024 | $0.01023 |
| Opus 5 | $0.00012 | $0.00511 |
| Sonnet 5 | $0.00005 | $0.00205 |
| Haiku 4.5 | $0.00002 | $0.00102 |
Grade A, and why
watch-upstream scanned grade A with 1 finding 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
If either tool is not found, fall back to `Bash(curl -s <url>)` for fetching. How it starts
The opening of the file, as written. The whole thing — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Watch Upstream
Detect changes in official Claude Code documentation that may require updates to dotforge.
Step 0: Discover tools
WebFetch and WebSearch are deferred tools — they may not be loaded yet.
Run ToolSearch("WebFetch WebSearch") to ensure both tools are available before proceeding.
If either tool is not found, fall back to Bash(curl -s <url>) for fetching.
Step 1: Fetch current documentation
Use WebFetch to read these pages directly:
https://code.claude.com/docs/en/overview— main feature listhttps://code.claude.com/docs/en/settings— settings.json schema, permissionshttps://code.claude.com/docs/en/hooks— hook types, events, matchershttps://code.claude.com/docs/en/memory— memory and context managementhttps://code.claude.com/docs/en/sub-agents— subagent capabilities (renamed from agent-tool)https://code.claude.com/docs/en/cli— CLI flags and options
If any URL fails, use WebSearch with query "Claude Code" <topic> site:code.claude.com as fallback.
If a URL returns 404 instead of redirecting, fall back to https://code.claude.com/docs/llms.txt (canonical index) and search for the slug — the new domain reorganized some paths (e.g. agent-tool → sub-agents).
Then search for recent announcements:
- WebSearch:
Claude Code new features 2026 - WebSearch:
Claude Code changelog site:anthropic.com - WebSearch:
Claude Code hooks settings update site:github.com/anthropics
Step 2: Extract and classify changes
For each finding, check if it affects dotforge:
| What to look for | Where it impacts dotforge |
|---|---|
| New hook event types (beyond PreToolUse/PostToolUse/Stop) | template/hooks/, stacks/*/hooks/ |
| New settings.json fields or changed schema | template/settings.json.tmpl, global/settings.json.tmpl |
| New permission categories | stacks/*/settings.json.partial |
| Changed deny list behavior | template/settings.json.tmpl deny section |
| New agent/subagent capabilities | agents/*.md |
| New skill/command system features | skills/*/SKILL.md |
| Deprecated features or breaking changes | Any affected file |
| New CLI flags relevant to automation | skills/benchmark/SKILL.md (uses claude --print) |
| MCP server changes | global/settings.json.tmpl |
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 · 108 lines · 24 tokens per session scan A 1f47e0b9bf89
watch-upstream is a skill published in the GitHub repository luiseiman/dotforge (8 stars, last pushed 2mo ago), licensed MIT. It adds 24 tokens to every session and 1,023 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
scaffold
Set up or align a non-code ops/PM repo with the ops folder structure, CLAUDE.md, and INDEX.md hierarchy.
find-opps
Find new directions for an ops project — adjacent opportunities, strategic gaps, preparatory research. Use when known work is well-mined and you need to look outward.
prioritize
Pick 2–3 independent ops tasks from active plans and backlog for autonomous execution, weighing impact, feasibility, and project goals.
factory-api
API conventions for both server actions and tRPC builds. Covers the decision between them, per-mutation Zod input schemas, central router composition, pagination shape, multi-field search via Drizzle ilike + or(), mutation lifecycle hooks, conditional query enabling, stale-time defaults, error response shape and…
factory-ci
CI and pull-request review conventions. One canonical .github/workflows/ci.yml is the merge gate (typecheck, lint, test, build, claude-review); branch protection's required-checks list matches the job list 1:1; anthropics/claude-code-action@v1 reviews every PR against the factory-pitfalls.md checklist as a required…
factory-design
Design-system discipline distilled across builds. Semantic token vocabulary (name intent, not palette position), CSS variables bridged into Tailwind, dark/light as a variable swap, single- vs two-layer token systems, component primitives as token consumers, when to promote a repeated utility cluster into a primitive…