agentleFS
ForeactLabsForeactLabs/OpenExecutive/.claude/skills/openexec-api/SKILL.md

openexec-api

Skill5.8k starsChanged 50 days agoApache-2.0View on GitHub

Interact with the Open Executive FastAPI backend via curl. Use this skill when the user asks to hit /chat, /today, /people, /scheduled_actions, /architecture/*, /health/*, /fixtures/*, /audit/*, or any HTTP endpoint on the Open Executive API, local or deployed. Authenticates via $BACKEND_SHARED_SECRET in the x-api-key header. Tiered safety: GET runs freely, mutating POSTs need explicit confirmation.

  • Reads credentials

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 1

    1. SenteLabsAI/OpenExecutive★ 5.8k100% same

    Edited · 0

    None found.

    Wrote their own version · 0

    None found.

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 0 are distinctive, and 5 are 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.

  • 5

    Very specific

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

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

    All recipes below use $OE_API.

    line 17
  2. Very specific:

    Never write the value to a file: Do not read from .env to "help" — if the env var isn't exported, ask the user to export it before …

    line 28
  3. Very specific:

    Always pipe JSON to jq so the transcript stays readable.

    line 139
  4. Very specific:

    If you need a route this skill doesn't list, the authoritative source is packages/core/openexecutive/api/routes/.

    line 145
  5. Very specific:

    After POST /fixtures/reset, the Honcho workspace is wiped.

    line 152

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

openexec-api

Curl recipes against the FastAPI backend — for hitting HTTP routes, not for inspecting SQLite or tailing logs on a host.

Base URLs

export OE_API=http://localhost:8000   # local (`make dev` must be running) — the default
export OE_API=https://<your-host>     # a deployed environment

All recipes below use $OE_API. If it is unset and the user doesn't specify, assume http://localhost:8000.

Auth

-H "x-api-key: $BACKEND_SHARED_SECRET"

Rules:

  • Never expand $BACKEND_SHARED_SECRET into the literal value in commands you print, in PRs, or in commit messages.
  • Never write the value to a file. Do not read from .env to "help" — if the env var isn't exported, ask the user to export it before proceeding.
  • For routes that vary by caller, optionally add -H "x-caller-email: <user's-own-email>". Never put someone else's email there — the backend uses that header to resolve the caller's Person row and pull their per-person Honcho memory.
  • Signed callers. When the API has CALLER_ASSERTION_PUBLIC_KEYS set, a bare x-caller-email is refused (401 caller_assertion_required), and a request with only x-api-key is a service that is never the owner. Owner-only routes refuse it. Then ask the user to export the UI's CALLER_ASSERTION_PRIVATE_KEY and sign each request instead of sending x-caller-email. Treat that key like the shared secret: never expand, print or store it.

    -H "x-caller-assertion: $(uv run --with cryptography python scripts/mint-caller-assertion.py GET /today)"

    Each assertion covers one request: the same method, the exact path and query, within 30 seconds, once. So mint one per call. It signs as the operator (the owner); --email <user's-own-email> signs as that user.

Safety tiers

  • Green — auto-run. All GET requests, and the cheap idempotent regenerate POSTs.
  • Yellow — confirm once per session. Writes that change durable state but are reversible (e.g. archiving a Person, deleting a scheduled action).
  • Red — print the curl, wait for explicit "go," no chaining. POST /fixtures/reset, POST /fixtures/{name}/load, POST /fixtures/unload, POST /fixtures/snapshot, POST /chat (impersonates the user — generally don't invoke from Claude).

Each red command needs its own "go" — a prior approval does not carry over.

Green: health & introspection

curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/health | jq
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/health/honcho | jq

/health/honcho returns one of:

  • {"status": "disabled"} — Honcho is off
  • {"status": "ok", "latency_ms": ...} — workspace reachable
  • {"status": "error", "error_type": "...", "error_msg": "..."} — wrapper or network failure

Green: architecture page

# All sections with freshness
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/architecture/sections | jq '.sections[] | {id, fresh, generated_at}'

# One section
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/architecture/sections/honcho | jq

# Cache key debug (per-component hashes feeding FactsBundle.core_hash)
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/architecture/debug/bundle | jq

Green/Yellow: force regenerate a section

Cheap, idempotent, but it does call Claude and write to disk — treat as yellow (one confirm per session is enough).

curl -sX POST -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/architecture/sections/honcho/regenerate | jq

Replace honcho with any section id from sections.py: overview, lifecycle, agents, caching, rag, review, memory, honcho, org, audit, schemas, workflows, scheduler, integrations, today, api.

Green: today / org / scheduler views

curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/today | jq
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/people | jq
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/people/by-scope/CEO | jq
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" "$OE_API/scheduled_actions?status=pending" | jq

Green: audit log

# Tail recent events
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" "$OE_API/audit/logs?limit=20" | jq

# Filter by event_type (peer_memory rows are Honcho ops)
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" "$OE_API/audit/logs?event_type=peer_memory&limit=10" | jq

# One event by id (returns full details_json)
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" "$OE_API/audit/logs/12345" | jq

Green: fixtures (demo companies)

# List available demos
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/fixtures | jq

# Current load state
curl -s -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/fixtures/status | jq

Red: fixtures (mutating)

Each of these is destructive or significantly state-altering. Print the exact curl, wait for "go," do not chain.

# Snapshot current local state before loading a demo (one-time setup before first demo load)
curl -sX POST -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/fixtures/snapshot

# Load a demo company (wipes current local state, switches Honcho workspace)
curl -sX POST -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/fixtures/<name>/load

# Unload demo (restores the snapshot, deletes the demo Honcho workspace)
curl -sX POST -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/fixtures/unload

# Full factory reset (wipes profile.yaml, company/, ChromaDB, ALL episodic tables,
# people store, departments, Honcho workspace, snapshot dir). Effectively a fresh install.
curl -sX POST -H "x-api-key: $BACKEND_SHARED_SECRET" $OE_API/fixtures/reset

Red: /chat

POST /chat runs the orchestrator end-to-end. Calling it from Claude would impersonate the user, log spurious turns to the audit trail, and (if x-caller-email matches the user) pollute their Honcho peer card. Document, don't invoke. If the user explicitly asks for a chat test from curl, treat as red and confirm.

Output handling

  • Always pipe JSON to jq so the transcript stays readable.
  • For long responses, project early: jq '.sections | length', jq '.[] | {id, status}', jq '.audit_events[0:3]'. Don't dump full bodies into Claude's context.
  • For text/streaming endpoints, redirect to a temp file and head/grep.

Endpoint discovery

If you need a route this skill doesn't list, the authoritative source is packages/core/openexecutive/api/routes/. One file per router. Grep for @router.get|@router.post|@router.delete|@router.patch to enumerate. Do not guess endpoints — the backend's middleware will 401 you on the wrong path with the api-key header.

Common gotchas

  • localhost:8000 requires make dev running. If curl returns "Connection refused," the dev server is down.
  • x-api-key middleware applies to nearly all routes. A bare GET /health works without it, but most others 401 without the header.
  • The UI proxies through /api/backend/... and re-stamps x-caller-email from the verified session. Hitting the API directly bypasses that — useful for testing, but means you control the caller identity (and have to set it accurately).
  • After POST /fixtures/reset, the Honcho workspace is wiped. The next chat turn rebuilds it lazily; don't expect prior conversation memory.

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.