| name | orchestration-depth1 |
| description | Short-form orchestration guidance for localized `depth=1` execution when token economy is active. Use for small, well-bounded tasks that do not require broader cross-module reasoning or mandatory depth escalation. |
| allowed-tools | ["Read","Grep","Glob","Bash(*)","Edit","Write"] |
| metadata | {"author":"garda-agent-orchestrator","version":"1.0.0","runtime_requirement":"Node.js 24 baseline for public CLI and gate commands"} |
Orchestration (Depth 1 Short Form)
Use this short form only when all of the following stay true:
- effective depth is
1;
- scope is small and well localized;
- no automatic escalation trigger fires;
- correctness does not depend on broad cross-module context.
Escalate back to the full orchestration skill immediately if:
- preflight reports
FULL_PATH;
- required
db, security, or refactor review forces depth>=2;
- the change touches auth, payments, sensitive data, infra, or other high-risk areas;
- scope drifts beyond the original task.
Minimal Required Inputs
- user request;
TASK.md row for the task;
AGENTS.md routing entrypoint;
- preflight artifact for the task;
- directly touched module context;
00-core.md and 80-task-workflow.md;
- only the rule ids/snippets directly triggered by changed scope.
Compact Workflow
- Start only from
Execute task <task-id> from TASK.md strictly through the orchestrator. Use `next-step` as the navigator; when independent review is required, launch a sub-agent using your internal tools. and show one English start marker from the repo-owned list (Garda captures my mind or Garda rewrites my code) once in the first relevant reply; this UX marker is not gate evidence. Do not use start-marker presence or exact text as hard evidence for task-mode, compile, review, completion, or task-audit gates.
- Capture requested/effective depth in non-status
TASK.md notes; successful enter-task-mode reconciles the task to IN_PROGRESS.
- Build a concise plan focused on changed files, risks, and validation.
- Run
enter-task-mode, load-rule-pack --stage TASK_ENTRY, handshake-diagnostics, shell-smoke-preflight, and then preflight/classification; stop using this short form if escalation is required.
- Read only minimal context: core rules, task workflow, touched module context, and scope-triggered rule snippets.
- Implement the smallest safe change that satisfies the task.
- Run objective validation for the touched area.
- Run the mandatory compile gate.
- Run only the required independent reviews from preflight; successful review-gate flow reconciles the task to
IN_REVIEW.
- Resolve findings, ensure the final PASS artifact has no active findings or residual risks unless accepted non-blocking items are moved to
Deferred Findings with Justification:, run completion gate, and only then let completion finalization reconcile the task to DONE.
Hard Rules
- Depth changes context budget, never gate obligations.
- Do not continue a normal task run when the workspace already had modified files before
enter-task-mode; isolate scope first with staged or explicit preflight inputs.
- Do not skip compile, review, or completion gates.
- Mandatory reviews still require a fresh clean-context delegated reviewer; do not reuse an existing reviewer session, and close or release the reviewer after receipt persistence.
- Do not hand-edit active
TASK.md lifecycle status transitions (IN_PROGRESS, IN_REVIEW, SPLIT_REQUIRED, DONE, BLOCKED); gate flow owns IN_PROGRESS, IN_REVIEW, SPLIT_REQUIRED, and DONE, explicit operator task-reset owns reset/discard, and DECOMPOSED is only for intentionally split parent tasks after child tasks are linked. Strict decomposition splits do not relax strictness: linked children must be parent-derived strict-profile rows that match the recorded decision artifact.
- Re-run preflight after meaningful scope changes.
- If the original preflight used planned
--changed-file inputs in a clean workspace before implementation, refresh it before compile by rerunning classify-change and then the POST_PREFLIGHT command printed by next-step once the real diff exists.
- Do not mark a task
DONE while any PASS review artifact still has active findings, residual risks, or deferred items without Justification:.
- Prefer concise evidence and scoped artifacts over pasting large raw outputs.
- The final summary must include the same output-compaction line shape as the full orchestration contract: chars first, approximate percentage when baseline is known, readable
+ separators between breakdown segments, and token estimate only as a secondary note when available.