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 instructions/sainteye/clawdline/agents-mdgit clone --depth 1 https://github.com/sainteye/clawdlineWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/sainteye/clawdline/agents-md)<a href="https://agentmods.dev/instructions/sainteye/clawdline/agents-md"><img src="https://agentmods.dev/badge/instructions/sainteye/clawdline/agents-md.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.10682 | $0.10682 |
| Opus 5 | $0.05341 | $0.05341 |
| Sonnet 5 | $0.02136 | $0.02136 |
| Haiku 4.5 | $0.01068 | $0.01068 |
Grade A, and why
clawdline AGENTS.md scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s "http://127.0.0.1:$PORT/v1/orchestrator/tasks" \ How it starts
The opening of the file, as written. The whole thing — 668 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Working in this repository
This checkout is shared by several live agent sessions, including Claude Code and Codex.
Treat every change already present in the opening git status as another session's unfinished work.
Do not stage, revert, rewrite, or commit those changes.
Do not infer ownership from a filename, a recent timestamp, or a new commit; inspect the content and scope.
Project vocabulary is defined without implementation detail in CONTEXT.md. Read it
before changing task, graph, review, landing, handoff, or verification semantics; follow its links
to the narrower operational document for the work in front of you.
Shared-tree discipline
- Record
git status --shortbefore editing so you know the pre-existing state. - Edit only paths assigned to your task.
- Use
git add -- <file>...with every path named explicitly. - Never use
git add -A,git add ., or another broad staging command. - Before any root-owned commit, run
git diff --cached --statand remove anything outside that commit's scope. What you are looking for is a path you did not stage yourself. The index is shared: another session'sgit addis already sitting in it, and a plaingit commitwith no pathspec commits the whole index, theirs included. That happened twice on 2026-08-28. Read the stat as "is every one of these lines mine?" rather than "does this look about right", and take a stranger back out withgit reset -- <path>, which unstages without touching their bytes. - Read the staged diff itself; a clean stat does not prove that a file contains only your work.
- After hunk-staging, commit the reviewed index with plain
git commit -m <message>and no pathspec.git commit -- <path>...takes the named files from the worktree instead of the staged hunk selection and can absorb another session's unstaged hunks. A commit pathspec is safe only when the entire worktree diff of every named path belongs to that commit. - The two commit traps point in opposite directions, which is why neither bullet above replaces
the other: no pathspec sweeps in what somebody else staged, a pathspec sweeps in what somebody
else left unstaged.
git diff --cached --statcatches the first, reading the staged diff catches the second, and a commit that has had neither run on it is guarded against neither. - A child or worker session does not commit. It hands its changes back to the root session for review and commit.
- Once a root session has completed the task it was originally assigned, reviewed its own diff, and verified the exact commit in isolation, it should commit its completed work without waiting for a separate user request to commit. This permission applies only when the root can isolate its own changes from every pre-existing or unrelated worktree change; otherwise it must keep the work uncommitted and coordinate with the owner of the overlapping paths.
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 · 668 lines · 10,682 tokens per session scan A f45619afd5c4
clawdline AGENTS.md is an instructions file published in the GitHub repository sainteye/clawdline (1 stars, last pushed 3d ago), licensed MIT. It adds 10,682 tokens to every session, about $0.0534 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 instructions, from other repositories
repoprompt-ce AGENTS.md
AGENTS.md instructions for repoprompt/repoprompt-ce, covering agent notes, contribution preflight, run, debug and uncoordinated equivalents.
diri AGENTS.md
AGENTS.md instructions for cristicretu/diri, covering diri repository instructions, scope, completed remote architecture baseline, rust workspace map and remote architecture invariants.
cctop AGENTS.md
AGENTS.md instructions for st0012/cctop, covering agents.md - development guide for cctop, scope and sources, required development rules, driver workflow and github and pr rules.
here-macos CLAUDE.md
Instructions for koalaauto/here-macos, covering here — ai assistant notes, architecture at a glance, prerequisites, build & run and always pass -quiet from an agent session.
signalbox CLAUDE.md
Claude Code instructions for dwmkerr/signalbox, covering signalbox, specs are the source of truth - keep them current, layout, build & test and regenerating the hero gif.
cctop CLAUDE.md
Claude Code instructions for st0012/cctop: This repository keeps shared agent instructions in AGENTS.md. Read AGENTS.md now and follow it as the source of truth for cctop development.