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/cintia09/codenook/remote-watchnpx skills add cintia09/CodeNook --skill remote-watchgit clone --depth 1 https://github.com/cintia09/CodeNookWhat 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 | $0.00002 | $0.01139 |
| Opus 5 | $0.00001 | $0.00570 |
| Sonnet 5 | $0.00000 | $0.00228 |
| Haiku 4.5 | $0.00000 | $0.00114 |
Grade A, and why
remote-watch 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
`subprocess.run(..., shell=True)`. The Python execution gives the How it starts
The opening of the file, as written. The whole thing — 114 lines — stays where its author put it; the contents beside it link to each section on GitHub.
remote-watch — plugin-shipped skill (development plugin)
Role
Probe a remote review / CI endpoint for the status of a submission and
return one of merged, pending, rejected, or unknown. The skill
itself knows nothing about Gerrit / GitHub / Jenkins / etc. —
specifics live in workspace memory or are supplied per-call.
Security / threat model
--config <path> is executed as a Python module (via
runpy.run_path); the module-level PROBE_CMD is then executed via
subprocess.run(..., shell=True). The Python execution gives the
config full host privileges (it can import os; os.system(...) etc.)
before any probe even starts. This is intentional — the memory-first
pattern needs full flexibility to wrap arbitrary review backends
(custom CLIs, multi-step queries, env-var preludes). Consequences:
- The caller is responsible for trust. Only point
--configat: (a) a workspace-memory entry under.codenook/memory/knowledge/that a human author wrote / reviewed; OR (b) a snippet just pasted by the current user via HITL. - Never load a
--configfile fetched over the network without human review first. - The skill makes no attempt to sandbox the config or the probe
— anything Python (and
shell=True) can do, the config can do. - Memory entries shipped by plugins (under
.codenook/plugins/<id>/) are not auto-trusted; they reach--configonly after a human promotes the embedded snippet to a memory file.
Three-tier resolution (memory-first → user-ask)
-
Tier 1 — cheap probe. If the target dir contains a
.github/folder andghis on PATH, rungh pr view <ref> --json state. If it contains a.gerrit/marker andsshto the host succeeds, run the recordedssh gerrit querycommand. These two probes ship as defaults so the most common cases just work. -
Tier 2 —
--config <path>. The caller (submitter role) does:<codenook> knowledge search "remote-watch-config target=<basename>"If a memory hit is found, the entry contains a Python snippet that defines (as module-level variables):
PROBE_CMD: str— command line to run; stdout becomes status textSTATUS_REGEX_MERGED: str— regex matched against stdoutSTATUS_REGEX_REJECTED: str— regex matched against stdoutSTATUS_REGEX_PENDING: str— regex matched against stdout (default".*")
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 114 lines · 2 tokens per session scan A 507de93eb79c
remote-watch is a skill published in the GitHub repository cintia09/CodeNook (5 stars, last pushed 3mo ago), licensed MIT. It adds 2 tokens to every session and 1,139 once invoked, about $0.0000 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
babysit
Same-session monitoring loop for PRs, CI runs, tickets, and deployments using the monitorstart / monitorupdate / autonudgestop MCP tools. The loop re-injects your check instructions into THIS session on an idle interval — same context, same tools — and works from dashboard chat, Slack threads, and Discord DMs. Use…
testing-and-ci
Testing conventions, CI pipeline rules, and smoke test coverage for SkillHub. Ensures agents write tests correctly and understand the CI gate requirements.
devops/deployment-process
部署流程和CI/CD配置，确保安全可靠的部署.
glab-cli
GitLab CLI (glab) reference and workflow for repository, merge request, issue, CI/CD, release, and API operations across GitLab.com and self-managed or dedicated instances. Use when Codex needs to run or explain glab commands, usually by relying on the current glab context first, and only falling back to git remote -v…
qa-ci-cd-testing
当需要把测试集成到 CI/CD 流水线中、或者现有流水线的测试环节跑起来效率低不可靠时使用此技能。覆盖流水线各阶段的分层测试卡点设计（提交检查→单元测试→接口测试→UI 测试→回归测试）、工具集成策略和质量门禁配置。不要在 CI 里堆满慢的 UI 测试——而是构建测试金字塔：提交阶段跑最快的（<5min），合码阶段跑核心的（<15min），夜间跑全量的。.
qa-test-automation-arch
当需要设计自动化测试框架、或者现有框架维护成本太高需要重构时使用此技能。运用 PageObject、分层测试、关键字驱动、数据驱动等模式设计可维护可扩展的自动化架构。不要直接写测试代码——先设计架构：选型（UI/API/单元）、分层（测试层/业务层/基础设施层）、数据管理（测试数据与脚本分离）和 CI 集成方案。好的自动化架构应该让写用例的人不需要懂底层实现。.