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 commands/sjarmak/coding-agent-workflows/project-initgit clone --depth 1 https://github.com/sjarmak/coding-agent-workflowsWhat 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 | $0.00027 | $0.00864 |
| Opus 5 | $0.00014 | $0.00432 |
| Sonnet 5 | $0.00005 | $0.00173 |
| Haiku 4.5 | $0.00003 | $0.00086 |
Grade A, and why
project-init scanned grade A with 0 findings 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Workflow: Project Init
Stand up the context layers for a project so an agent has the right knowledge
in the right place from day one. The output is a thin, pointer-style AGENTS.md
(intention + an empty failure-mode log + references) and seed COMPASS.md maps
for the largest areas. It deliberately does not write the memory layer
(CLAUDE.md / instincts) — see context-layering for why each fact gets one home.
This is a runtime-neutral spec. Any orchestrator can drive it; the steps that require judgment are delegated to the model (ZFC).
When to run
- Setting up a new or existing repo for agent-assisted work
- The user says "set up AGENTS.md" / "initialize agent context" / "set up compass"
- Adopting this bundle in a project that has no intention layer yet
Steps
1. Reconnaissance
Run codebase-onboarding Phases 1-3 to detect stack, structure, entry points,
and conventions. This is shared signal for both outputs below. Do not yet write
any file.
2. Scaffold the thin AGENTS.md
From the recon, instantiate the AGENTS.project.md template at the repo root:
- What this project is — fill intention: the invariants and non-obvious constraints a capable agent would otherwise get wrong. Not a README restatement.
- Failure-mode preventions — leave empty with its format note; entries arrive
later via
failure-mode-capture. - References — point
{{BUNDLE_REF}}at where the bundle was installed (./.claudefor Claude Code, the bundle'sAGENTS.md/rulesotherwise),{{PROJECT_DOCS}}at the repo's own docs.
If an AGENTS.md already exists, enhance, don't overwrite: preserve its
content, slot existing project rules into the right sections, and call out what
was added. Honor the boundary — move anything that is really area-detail into a
compass file (step 3) and anything host-specific toward memory.
Keep it under the ~120-line target. If recon produced more than fits, that is signal the excess belongs in compass files, not AGENTS.md.
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 · 78 lines · 27 tokens per session scan A e4b03d9c7f92
project-init is a command published in the GitHub repository sjarmak/coding-agent-workflows (2 stars, last pushed 1mo ago), licensed MIT. It adds 27 tokens to every session and 864 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
project
Manage projects: register repos, inspect, configure per-project settings, and remove.
import
Import reads the legacy Agent Orchestrator flat-file store (/.agent-orchestrator) read-only and ports its projects and per-project settings into the rewrite database. Legacy files are never modified, and a re-run skips rows that already exist, so it is safe to run more than once. The daemon must be stopped before…
pm-all
Full technical governance scan; rebuild .pm/dashboard. Default does not wait on draft PRD. Use --compare-baseline only after a confirmed PRD/charter.
dispatch-worker
Run one in-session-agent Worker tick (RFC-0041 §4.3.1).
pm-done
Close a todo (TODO-xxx), sync completed.md, refresh overview.
pm-outline
Generate detailed project outline and draft charter from user intent (empty/new projects).