agentleFS
williamcorrea23williamcorrea23/sap-router-skill/.agents/skills/arc-1-skills/SKILL.md

arc-1-skills

Skill1 starsChanged 4 months agoMITView on GitHub

ARC-1 agentic SAP development skills (arc-mcp/arc-1/skills). Reference for ADT tools, code activation, and automated ABAP editing. Covers the 8 arc-1 MCP tools (SAPContext, SAPDiagnose, SAPLint, SAPManage, SAPNavigate, SAPRead, SAPSearch, SAPTransport), the read-edit-activate-transport flow, and arc-1's role as PRIMARY ADT MCP in the routing tree.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

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

    If arc-1 is down, the TieredFallback engine (scripts/fallback_engine.py) cascades ADT -> RFC -> GUI -> BDC -> Offline -> Manual.

    line 47

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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
SKILL.md

ARC-1 — Primary ADT MCP for Agentic ABAP Development

arc-1 is the PRIMARY ADT MCP of this orchestrator. All ADT-class operations (read source, edit, activate, lint, transport) route to arc-1 first, then aibap (secondary), then mcp-sap-gui (fallback). See scripts/sap_router.py (SapRouter.get_route()).

1. Tool Map — When to Use Which

ToolUse for
SAPContextSystem/session context — SID, client, user, release. Call first in a new session
SAPSearchFind objects by name/type when you do not know the exact object name
SAPNavigateJump between related objects — includes, super/subclasses, where-used
SAPReadRead source code and metadata of a known object (class, program, CDS, table)
SAPLintStatic checks (syntax, ATC-style) on source BEFORE activation
SAPManageCreate / modify / activate / delete objects — all write operations
SAPDiagnoseShort dumps (ST22-style), traces, runtime diagnostics after failures
SAPTransportList / create / assign / release transport requests

Rule of thumb: read-only questions -> SAPSearch/SAPRead/SAPNavigate. Changes -> SAPLint -> SAPManage -> SAPTransport. Failures -> SAPDiagnose.

2. Route Through the Orchestrator

# Let the router pick arc-1 (ADT-first) for the action:
npm run router -- --action ADT_READ_SOURCE
python scripts/sap_router.py route --action ADT_EDIT_CLASS

# Force GUI fallback only when ADT cannot do it (SPRO, dynpro):
npm run router:gui -- --action SPRO_CONFIG

If arc-1 is down, the TieredFallback engine (scripts/fallback_engine.py) cascades ADT -> RFC -> GUI -> BDC -> Offline -> Manual. Do not hand-code fallbacks.

3. Typical Flow: Read -> Edit -> Activate -> Transport

  1. Context — SAPContext to confirm system/client (never assume DEV).
  2. Locate — SAPSearch for the object; SAPNavigate to find the exact include/method.
  3. Read — SAPRead to get current source. Edit surgically (Karpathy: touch only what's needed).
  4. Lint — SAPLint on the modified source. Fix findings before writing.
  5. Write — SAPManage to save the change (object goes INACTIVE).
  6. Activate — SAPManage activate. An inactive object is invisible to runtime.
  7. Transport — SAPTransport to assign the object to a modifiable request. Release only at gate (npm run abap:review:transport).
  8. Verify — SAPRead the active version back; SAPDiagnose if execution dumps.

4. Health and Learning

# Probe arc-1 (HIGH criticality) + .env completeness:
npm run hc
npm run hc:prompt      # interactive fix for missing SAP_HOST/SAP_USER/SAP_PASSWORD

# Feed routing telemetry so self_learn.py keeps arc-1 primary:
npm run learn:mcp -- --mcp arc-1 --latency 245 --success true
npm run learn:route -- --action ADT_EDIT_CLASS --success true

Pitfalls

  • 401/403 on every arc-1 call → Cause: SAP credentials missing or wrong in .env, or user lacks S_DEVELOP. Solution: run npm run hc:prompt to fill SAP host/user/password; confirm developer role with Basis.
  • "Object locked by user X" → Cause: object is edit-locked in SE80/ADT by another session (or a stale lock of your own). Solution: retry later or ask the lock owner to release (SM12); never force-delete locks on shared objects.
  • Change saved but behavior unchanged → Cause: object left INACTIVE — SAPManage save without activate. Solution: always activate after write and re-read the active version to confirm.
  • Activation fails with syntax errors → Cause: edit broke dependent objects or lint step was skipped. Solution: run SAPLint before SAPManage write; use SAPNavigate where-used to check dependents; SAPDiagnose for details.
  • "Not assigned to a transport" on save → Cause: object in a transportable package but no modifiable request. Solution: SAPTransport create/assign a workbench request first, or use $TMP for throwaway objects.
  • arc-1 timeout / connection refused → Cause: MCP server down or SAP host unreachable. Solution: npm run hc to confirm; router auto-falls back to aibap then GUI — do not bypass the fallback engine.

Verification

# 1. arc-1 healthy and env complete:
npm run hc

# 2. Router resolves an ADT action to arc-1 (primary):
python scripts/sap_router.py route --action ADT_READ_SOURCE

# 3. After an edit: SAPRead the object and confirm version = active,
#    then gate before release:
npm run abap:review:transport

Related

  • sap-router-skill — master routing tree (ADT -> arc-1 first, GUI fallback)
  • abap-code-review — 9-dimension GO/NO-GO gate before SAPTransport release
  • sap-transport-gate / sap-transport-management — transport landscape rules
  • karpathy-guidelines — Think, Simplify, Surgical, Goal-Verify wrapper for every edit

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.