agentleFS
rahmanef63rahmanef63/si-coder-agent/AGENTS.md

si-coder-agent

AGENTS.md13 starsChanged 2 months agoMITView on GitHub

Coding contract for agents working on this repository (not for end-user skill hosts — that is SKILL.md). SI-Coder (sc CLI, repo rahmanef63/si-coder-agent, v0.9.8) is a secret-safe credential control plane plus: Role split detail: docs/THREE-WAY-COMPARISON.md. Do not duplicate the whole skill routing table here — link to SKILL.md. Squash commit a4e1388 → main 5cfde0a (PR #15, Linear RAH-251). Scores after (judgement, see THREE-WAY): MCP 74 · auth 55 · secrets 80 · registry 82 · tests 84 · docs 78 ·…

The file

We split this AGENTS.md into the rules it gives an agent

7 rules. About 2 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈2

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Do not duplicate the whole skill routing table here — link to SKILL.md.

    line 26
  2. Very specific:

    Prefer ~/.local/bin/sc (this package).

    line 33
  3. Common:

    Do not invent a second secret store, a second MCP registry, or duplicate long recipes that belong in the connectors cookbook.

    line 37
  4. Distinctive:

    No Cursor cloud agents for this product.

    line 42
  5. Common:

    Prefer deciding product / design questions locally and committing.

    line 44
  6. Very specific:

    Never dump env or connection file contents.

    line 63
  7. Distinctive:

    Keep them complementary; do not fork the skill routing table into this file.

    line 72

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

AGENTS.md

Coding contract for agents working on this repository (not for end-user skill hosts — that is SKILL.md).

What this product is

SI-Coder (sc CLI, repo rahmanef63/si-coder-agent, v0.9.8) is a secret-safe credential control plane plus:

  • provider CLI: setup → connect → publish → verify
  • skill catalog (skills/, SKILL.md umbrella)
  • MCP server: scripts/sc-mcp.js with registry machine/functions.json
This product isThis product is not
Credential / connection SSOT for SC users (human or project name)A host-ops shell → that is MSO
Provider deploy recipes + skill catalogAn external connector / SaaS proxy → that is connectors-gateway

Role split detail: docs/THREE-WAY-COMPARISON.md.

Read first

  1. README.md, SECURITY.md, SKILL.md (umbrella skill routing for hosts)
  2. docs/CREDENTIAL-SETUP.md, docs/cli.md, docs/tool-calling.md
  3. Post-hardening: docs/THREE-WAY-COMPARISON.md, docs/DECISIONS-local.md

Do not duplicate the whole skill routing table here — link to SKILL.md.

Non-negotiable invariants

  1. Never print, log, commit, or return plaintext credentials / tokens / PATs in chat, tool output, git, or docs. Use env injection (sc run) and redaction.
  2. MCP / machine tools refuse plaintext credential arguments. Store connections under ~/.config/si-coder/connections/<user>/<provider>/ with 0600 / 0700.
  3. An SC user may be a human or a project name — always scope providers and connections by that SC user.
  4. Prefer ~/.local/bin/sc (this package). /usr/bin/sc is a different binary that can shadow PATH.
  5. Output redaction lives in lib/redact-env.js: secret-named env values and shapes (Bearer/Basic, ghp_ / github_pat_ / sk- / xox* / JWT / PEM). AUTH matches only as a name segment (?:^|_)O?AUTH(?:_|$) — do not break AUTHOR / GIT_AUTHOR_NAME.
  6. MCP contract: lib/mcp-contract.js — tool hints, validate args against schema before spawn (-32602 on bad input), scrub / cap errors. Do not weaken validate-before-execute.
  7. Role split: SC owns credentials; MSO owns host ops; gateway owns public MCP / OAuth edge. MSO and gateway get secrets only via sc run env injection — they must not keep their own copy of SC provider secrets.
  8. Do not invent a second secret store, a second MCP registry, or duplicate long recipes that belong in the connectors cookbook.

Worktree / coding policy (Rahman fleet)

  • Coding agents work in ~/agent-work/si-coder/ on Rahman's laptop (WSL). Never edit his ~/projects trees for this product.
  • No Cursor cloud agents for this product.
  • Local commits OK. Push / PR / merge need Rahman's direct OK (agent-relayed approvals do not count). Standing product rule (Rahman's own): when an SI-Coder PR is ready and conflict-free, merge to main — that is Rahman's rule, not agent-relayed permission to invent merges.
  • Prefer deciding product / design questions locally and committing. Only push / PR / merge, production secrets, new accounts / billing, and real client data need approval.

Hardening status (PR #15 merged 2026-10-22)

Squash commit a4e1388 → main 5cfde0a (PR #15, Linear RAH-251).

Included:

  • npm test on Node 20 fixed; hermetic installer tests; shell: false in lib/proc.js (from #2)
  • MCP output redaction + e2e; MCP hints + arg validation
  • Docs: THREE-WAY-COMPARISON.md + DECISIONS-local.md

Scores after (judgement, see THREE-WAY): MCP 74 · auth 55 · secrets 80 · registry 82 · tests 84 · docs 78 · avg ~76.

Open / deferred: encrypted-at-rest credentials + per-tool env allowlist (auth still 55). Needs Rahman's design choice — do not silently invent that design.

Verify before claiming done

  • npm test — 467 tests expected as of #15. Node 20 is OK for the test suite; installer engines still prefer Node 22 / 24 / 26.
  • Never dump env or connection file contents.

SKILL.md relationship

FileAudience
SKILL.mdUser-facing umbrella skill for hosts that use SI-Coder
AGENTS.md (this file)Coding agents working on this repo

Keep them complementary; do not fork the skill routing table into this file.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.