Use native Codex model, effort and agent selection. Respect the operator's selected main model and effort; this skill does not switch an active task. The profiles below are policy recommendations. They do not establish the current host configuration or model selected by a running task; check those separately before describing them as configured. Choose the execution arrangement before a delegated worker's model. Authorization to delegate alone does not require consulting this router.
Execution arrangement
Keep small, tightly coupled work and immediate follow-ups with the agent that already holds the relevant context. A cheaper worker model alone does not justify a handoff. Batch related operations rather than delegating each one.
When the operator requests delegation, preserve that workflow for suitable bounded work. Otherwise delegate when the benefit justifies briefing, startup, coordination, verification and likely correction costs. Reuse a suitable worker and its relevant context for related follow-ups, without assuming cache reuse. Give it the context needed for its acceptance checks.
Parallelise independently executable packages with clear ownership when the benefit outweighs duplicated context and integration costs. Sequence dependent work. Faster completion and lower consumption are separate benefits.
Continue the current thread when its context remains useful. A fresh task suits an independent objective; a full-history fork copies history rather than providing a clean context.
Keep one visible task per coherent outcome. Use native subagents for useful, bounded work inside that task. Create a separate user-owned task only on the operator's explicit request, for an independent or durable outcome, a schedule, or a required host/cloud boundary.
Operator profiles
| Profile | Model | Effort | Use |
|---|---|---|---|
| Recommended coordination profile | gpt-6.1-sol |
medium |
Recommended for sustained coordination, integration and open-ended execution when available; verify the actual host selection separately. |
| Planning, review and difficult decisions | gpt-6-astra |
low |
Habitual Astra setting when its judgment helps. |
| Exceptional reasoning | gpt-6-astra |
xhigh |
Rare difficult decisions or unresolved investigations after the task warrants deeper effort. |
| Bounded subtask | gpt-6-luna |
high |
Narrow, well-specified work with clear ownership and acceptance checks. |
| Defined execution | gpt-6-luna |
xhigh |
Implementation or transformation with a concrete outcome and verification. |
| Substantial execution | gpt-6-luna |
max |
Justified when deeper effort materially helps a bounded implementation with observable proof. |
For demanding coding, GPT-6.1 Sol high is an available task-specific override when medium is insufficient; choose it for the work rather than forcing an Astra escalation. It does not replace the recommended medium coordination profile.
Choose directly for the task. Use an Astra → Sol → Luna sequence only when coordination materially helps the outcome; a direct bounded worker or the current agent is enough when it does not. There is no mandatory model ladder, failed-attempt retry, coordinator stage, automatic planning/review delegation, CLI resolver, receipt workflow or mandatory preset wrapper. Requested planning or review delegation remains available when suitable. Importance alone does not justify XHigh. A read-only task can still need Astra. Patch size alone does not prove that Luna suits autonomous systems administration.
Luna Max is available when justified by the work; it requires no repeated approval merely for that effort. Choose effort for the workload; higher effort does not guarantee better results. GPT-6 supports cache-preserving effort updates through its documented mechanism; do not assume every client uses it or that cache is shared across models or tasks. Current GPT-6.1 capability, pricing and benchmark evidence is in references/gpt61-sol-20260929.md. The historical GPT-6 evidence remains in references/gpt6-20260922.md. These profiles do not authorise new spending routes, external actions or additional access. Astra Max and Ultra are outside the habitual palette. Terra and other supported models remain available when selected by the operator or bound to a specialist. Honour explicit selections and fixed bindings, including those outside this palette.
For pstack's three role-aware presets, use the single canonical mapping in references/pstack-role-presets.json. Equilibrado is the default for unbound roles in a new setup with no existing role choices. Economia, Equilibrado and Power preserve explicit model/effort choices, panel lists, aliases, named specialists and the operator's selected main model/effort. Show a preserved binding that differs from a preset as an explicit override. If a requested model or effort is unavailable on its channel, keep it unchanged and mark it as needing a choice. Select either a role-aware preset or one upstream global budget; never apply the global effort target over a role-aware preset.
Changing selection during work
Keep GPT-6.1 Sol medium as the recommended policy default and choose
task-specific overrides through the native surface that actually supports them.
Honor an explicit native effort selection, including low; this recommendation
does not require changing a home configuration or an active thread. In CLI,
/model changes model and
supported effort within a session. A task follow-up tool may expose model and
effort overrides; use them only when that tool supports the target. Do not
message the current task recursively to simulate self-switching. The current
agent cannot claim to change its own model without an exposed control and
confirmation. Use the existing bounded specialist delegation when suitable.
Changing config.toml sets a default, not proof that an active turn switched.
Runtime and delegation
Before selecting compute, check the exact channel's current advertised models and efforts. Current-task, native-subagent and user-owned-task support differ. Do not invent unsupported capability, create a task to work around it, or silently substitute. Report the gap and continue locally when the existing selection can meet the proof bar; otherwise surface the specific decision needed.
For permitted delegation, give bounded ownership, relevant context, expected output and verification requirements. A differently modelled native worker needs fresh or bounded context: full-history forks inherit the parent model. Respect actual available slots; a concurrency ceiling is neither a target nor permission to spawn. Do not make specialists coordinators or enable recursive delegation implicitly. The coordinator integrates and verifies the result.
Creating a user-owned task, forking, moving or archiving tasks still requires the corresponding explicit user request. Routing grants no authority over hosts, accounts, private data, credentials, publication or destructive actions. Respect native approvals and refusals. Do not force feature flags.
Before accepting a delegated deliverable, check the requested elements against the actual result and verification evidence. Concise reporting must not remove required work or presentation quality. Correct omissions before delivery; change the assignment or model when observed failures justify it.
Cost and evidence
Compare the remaining whole task, including parent and worker input, output, briefing, startup, coordination, verification and rework. Reusing a relevant worker can save context transfer; cached input is possible, not guaranteed across tasks, models or changed prefixes. Do not add speculative cache retention, keepalive prompts or repetitive polling merely to preserve context or show activity. Count reasoning and child usage once if a meter already includes them. API token prices and benchmarks inform comparisons but do not measure this account's Codex allowance or prove a routing saving. Use observed account usage or a like-for-like native task comparison for such claims; state when that evidence is unavailable. Keep changing prices and benchmark scores out of durable profile rules.
Installation and updates
The sole install source is TheAngryPit/TheAngrySkills, not Workbench.
npx skills add https://github.com/TheAngryPit/TheAngrySkills.git -g -a codex -y -s model-capability-router
Use one canonical installed skill per OS user, shared by Codex profiles through
native discovery or links. A profile link is not a second independent version.
Keep source metadata pointed at the normal repository. npx skills records
update provenance; it does not provide an automatic update schedule. Inspect
the installed CLI's update/check behavior before using it as a read-only check.
Native named-agent profiles
This skill is also the public distribution surface for the native named-agent profiles used by the adapted pstack workflows:
assets/agents/comment-sicko.tomlis the named reviewer forcursor-no-comments.assets/agents/poteto-agent.tomlis the named delegate forcursor-poteto-modeplaybooks.
Keep these pstack specialist identities, ownership and verification contracts when the profiles are used; model routing does not flatten named roles.
The assets are not loaded merely because this skill is installed. To make the profiles selectable by a new Codex session, run the bundled checked installer from this skill directory:
python3 scripts/check-native-agent-profiles.py \
--install --installed-dir ~/.codex/agents
The bundled script travels with an individual skill installation; it does not
depend on the repository-level scripts/ directory.
The command refuses to replace a differing existing profile unless
--replace is supplied explicitly. Run it again with only
--installed-dir ~/.codex/agents for a byte-for-byte check. This validates
distribution and configuration, not that a newly started session has loaded or
selected the profile; that must be checked by the host session.
Additional Cursor source profiles
The profile assets also preserve Dylan, Advisor, compatibility review, continual-learning, plugin architecture, CI watching, and Thermos roles. sources/cursor-plugins/agent-coverage.json in the source repository records all 14 upstream agent mappings and their hashes. Source distribution is not installation or fresh-session loading. The existing checked installer above selects the two Pstack profiles; other assets remain separately selectable source profiles and must be explicitly installed and proven on the target host. Missing exact native types are reported, never claimed from these files alone.