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/zfinix/aster/cli-toolboxnpx skills add Zfinix/aster --skill cli-toolboxgit clone --depth 1 https://github.com/Zfinix/asterWhat 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 | $0.00055 | $0.01002 |
| Opus 5 | $0.00028 | $0.00501 |
| Sonnet 5 | $0.00011 | $0.00200 |
| Haiku 4.5 | $0.00006 | $0.00100 |
Grade A, and why
cli-toolbox 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 2d 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.
`curl -s -o /dev/null -w "%{http_code}" URL`. Content check: What it actually says
CLI toolbox
-
Read the tool before guessing at it, and never guess twice. First contact with a CLI you have not used this session:
command -v <tool>to confirm it is installed. When a call fails on usage rather than on the task ("unknown flag", "accepts at most 1 arg(s)", "unknown command"), the next call is<tool> --help, or<tool> <subcommand> --helpfor the exact subcommand that failed. Never re-send a shape that just failed with a small edit; that is how a turn burns twenty calls and lands nowhere. When the tool is missing, or its help does not settle the question, go and read:aster_mcpwithaction: "execute",name: "web/search"for the flag, orname: "web/extract"on the tool's documentation URL. Going online once costs less than four blind retries, and unlike them it ends the question. Say so plainly if the tool simply is not installed, rather than quietly substituting a different one. -
Ask the tool for structured output. Most modern CLIs have it:
gh pr view 12 --json title,files,cargo metadata --format-version 1,docker ps --format json,npm ls --json. Parse withjq -r '.field'instead of eyeballing prose. -
Slice files without reading them whole.
sed -n '120,160p' filefor a line range,grep -n -m5 "pattern" filefor first hits with line numbers,wc -l filebefore deciding how to read it. -
Count and rank instead of scrolling.
grep -cfor how many,sort | uniq -c | sort -rn | headfor what dominates,du -sh */ | sort -rh | headfor what is big. -
Probe HTTP minimally. Status only:
curl -s -o /dev/null -w "%{http_code}" URL. Content check:curl -s URL | grep -c "expected". Never dump a whole page to look at one thing. -
Force non-interactive mode. Anything that might prompt hangs the call: use
--yes/-y,--forcewhere safe,GIT_TERMINAL_PROMPT=0,DEBIAN_FRONTEND=noninteractive,CI=1,</dev/nullas a last resort. Anything that might page gets--no-pageror| cat. -
Quiet flags cut noise at the source.
npm install --silent,cargo build -q,git -qvariants. Prefer them over filtering noise afterwards. -
Batch transformations with xargs, not loops of calls.
grep -rl "old_name" src | head -20to see the blast radius, then one edit per file, not one search per file. -
Never
sed -ia repository file. Useedit_file: it is exact, previewed, and audited.sed -ialso differs by platform (BSD needssed -i '', GNU takessed -i), which makes it a portability bug on top of an audit hole. Stream edits (sed -n '10,40p',sed 's/x/y/'in a pipe) are fine anywhere. -
Check the platform line before GNU-only flags. The environment note says macos or linux. On macos the stock tools are BSD: no
grep -P(usegrep -E), differentdatearithmetic (-v+1d, not-d "+1 day"), noxargs -d. When a one-liner needs GNU behavior, reshape it around portable flags instead of assuming Linux. -
On Windows there is no sed or grep. Stock tools are PowerShell:
Select-Stringfor grep,Get-Content -Tail 20for tail. Ask the environment note, not habit, which world you are in.
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.
- 2d ago First seen · 59 lines · 0 tokens per session scan A aa2762677218
cli-toolbox is a skill published in the GitHub repository Zfinix/aster (40 stars, last pushed 7d ago), licensed Apache-2.0. It adds 55 tokens to every session and 1,002 once invoked, about $0.0003 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 2026-08-30.
Other skills, from other repositories
new-plugin
Factory line for adding a new HAR verification plugin (like playwright or rocketsim) for any framework — research the framework docs, build the template under src/templates/plugins/, register it everywhere, validate on a real repository, and open a PR. Use when asked to add/create a plugin, plugin template, or…
factory-line
Factory line for executing one station of a declared multi-station program — read the installed line bundle (har line status), plan parallel work into isolated HAR slots, run the cumulative gate with har line gate, and hand off for human review. Use when asked to "run a factory line", "run the next station", "execute…
v1-milestone
Factory line for executing one milestone of the HAR v1.0.0 refactor (epic os-factory/har#225) — plan the wave of parallel subagents, implement each issue in its own HAR slot, ship stacked PRs, run the fixture-e2e milestone gate, and hand off for review. Use when asked to "run the next v1 milestone", "work on v1.0.0"…
lemma-user
Operate an existing Lemma pod from the CLI as a human or agent: inspect resources, query tables and records under RLS, search and read pod files (converted markdown, page images), run functions and workflows, submit waiting workflow forms, chat with pod agents, run first-party tools, and execute third-party connector…
lemma-builder
Design and build complete Lemma pods: model tables/files/functions/agents/workflows/schedules/connectors/surfaces/apps from a problem statement, author a local pod bundle, import progressively with the Lemma CLI, and verify every layer. Use for pod design, creation, restructuring, import/export, and app development.…
lemma-research
Run rigorous, source-backed research in an existing Lemma pod. Use for investigations, literature or market scans, policy and product research, fact-checking, comparisons, current-information questions, and updates to prior research that require pod files, workspace material, or web sources; exact citations; explicit…