Orch infra
Most AI coding tools help you write code. Siegard Code manages the entire development lifecycle — it writes specifications, plans backlogs, implements features, runs QA, and delivers tested code. All autonomously, all traceable, all through Claude Code.
npx -y skills add zig999/siegard-code --skill orch-infraAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 9 stars9 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Shared infrastructure checks run at the start of every orchestrator cycle — run_preflight.py, run_integrity.py, run_circuit_check.py. Any non-zero exit blocks the cycle with status blocked. Loaded by all orchestrators. Not user-invocable — orchestrators run the scripts directly.
SKILL.md
3.5 KB, as published. Nobody here has run it
orch-infra
Shared infrastructure skill: run preflight, integrity, and circuit-breaker checks at the start of every orchestrator cycle.
Usage contract
Every orchestrator loads this skill and calls all three scripts at the start of each cycle:
python3 .claude/skills/orch-infra/scripts/run_preflight.py
python3 .claude/skills/orch-infra/scripts/run_integrity.py
python3 .claude/skills/orch-infra/scripts/run_circuit_check.py
On any non-zero exit: stop and return {"status": "blocked"} to the meta-orchestrator.
Output envelope (all scripts)
Every script emits exactly one JSON object to stdout:
| Field | Type | Description |
|---|---|---|
status | "ok" | "blocked" | Pass/fail verdict |
check | string | "preflight" | "integrity" | "circuit" |
timestamp | string | ISO 8601 UTC timestamp |
reason | string | Present only on "blocked" — machine-readable failure code |
detail | object | Present only on "blocked" — additional context |
scripts/run_preflight.py
Runs preflight.py --quick (local runtime checks only). Returns blocked if any check fails.
python3 .claude/skills/orch-infra/scripts/run_preflight.py
Exit codes: 0 = all checks passed, 1 = at least one check failed or exec error.
Example output (ok):
{"status": "ok", "check": "preflight", "timestamp": "...", "passed": 5, "total": 5, "failed_count": 0}
Example output (blocked):
{
"status": "blocked",
"check": "preflight",
"timestamp": "...",
"reason": "preflight_failed",
"passed": 4,
"total": 5,
"failed_count": 1,
"failed_checks": [{"check": "flock_works", "reason": "fcntl not available"}]
}
scripts/run_integrity.py
Verifies the hash chain of the orchestration log in strict mode. Returns ok if no log exists (first run).
python3 .claude/skills/orch-infra/scripts/run_integrity.py
Exit codes: 0 = chain intact (or no log), 1 = chain invalid or error.
Example output (ok):
{"status": "ok", "check": "integrity", "timestamp": "...", "events_verified": 42}
Example output (blocked):
{
"status": "blocked",
"check": "integrity",
"timestamp": "...",
"reason": "chain_invalid",
"events_verified": 41,
"first_error_seq": 42,
"truncation_candidate": 42
}
scripts/run_circuit_check.py
Evaluates circuit breaker state. Returns blocked if the circuit is tripped or should trip. Returns ok if no log exists (first run).
python3 .claude/skills/orch-infra/scripts/run_circuit_check.py
Exit codes: 0 = circuit open (healthy), 1 = circuit tripped or evaluation error.
Example output (ok):
{"status": "ok", "check": "circuit", "timestamp": "...", "tripped": false, "failure_count": 2, "threshold": 50}
Example output (blocked):
{
"status": "blocked",
"check": "circuit",
"timestamp": "...",
"reason": "circuit_tripped",
"tripped": true,
"failure_count": 52,
"threshold": 50
}
What orch-infra does NOT do
- Does not know about phases or domains
- Does not read spec files or delivery artifacts
- Does not interact with humans
- Does not emit events to the log