Using AI Skills

Astraeus bundles ten user-facing skills for assistants that support the Agent Skills format. They help turn an engineering question into a valid input, execute the installed product, and interpret its saved evidence without relying on repository access or developer tooling.

flowchart LR QUESTION["Engineering question"] --> COMPOSE["Draft or adapt input"] COMPOSE --> RUN["Validate and execute"] RUN --> EVIDENCE["Astraeus artifacts"] EVIDENCE --> REVIEW["Review, compare, or report"]

Skills are optional workflow assistants. The CLI, RunSpec, saved artifacts, reviews, and reports remain usable and reproducible without an AI model.

Install And Verify

List the skills included with the installed product:

astraeus skills list

Install the complete set for one supported assistant:

astraeus skills install --all --target codex --scope user
astraeus skills verify --all --target codex --scope user

Installation layouts are available for codex, agents, gemini, and claude. The target selects the directory and adapter format; all four layouts can be installed and verified from the same bundled skill payloads. End-to-end workflow qualification covers Codex, Claude Code, and Antigravity CLI using the generic Agent Skills layout. Copilot CLI also discovers that generic layout.

Use --scope project when the skills should apply only inside the current project, or install one skill by name instead of using --all:

astraeus skills install astraeus-runspec-drafter --target codex --scope project

Verification compares the installed skill payload with the version bundled in Astraeus. Use --force on a later install when you intentionally want to replace an edited or older copy.

Skill discovery and payload verification confirm installation. Assistant behavior also depends on the selected host, model, permissions, and execution environment, so validate important workflows in the environment where they will be used.

The Ten User Skills

Workflow	Skill	Use it for
Compose	astraeus-runspec-drafter	Draft or adapt a v8 RunSpec from the bundled examples and validate it.
Execute	astraeus-runner	Validate and run one RunSpec, report progress, and optionally launch the Console.
Review	astraeus-run-reviewer	Produce and explain a deterministic review of one completed core run.
Study	astraeus-study-manager	Plan, execute, resume, or rerun sweeps and batches.
Study	astraeus-study-reporter	Compare completed sweep or batch cases and identify material differences.
Campaign	astraeus-campaign-composer	Draft or adapt a constellation-deployment campaign request.
Campaign	astraeus-campaign-runner	Validate and execute a campaign, then report status and artifact locations.
Campaign	astraeus-campaign-analyzer	Interpret campaign closure, assignments, residuals, and mission burden.
Report	astraeus-scenario-reporter	Generate and explain a decision-level scenario report and plot pack.
Operations	astraeus-operations-manager	Check an installed release and operate already-qualified server workflows.

astraeus skills install installs these ten user-facing workflows. Repository maintenance, source-code changes, internal mission planning, and development-environment management remain repository tooling for maintainers.

Core Run Workflow

flowchart LR A["Describe the scenario"] --> B["runspec-drafter"] B --> C["runner"] C --> D["runs/core/run-id"] D --> E["run-reviewer or scenario-reporter"] D --> F["Astraeus Console"]

Example prompts:

$astraeus-runspec-drafter Adapt the regional-coverage example for
Scandinavia, keep the represented-surface aggregation, and validate the result.

$astraeus-runner Run my-regional-coverage.yaml with run id scandinavia-study,
report progress, and open the Console when it completes.

$astraeus-scenario-reporter Create a decision-level report for
runs/core/scandinavia-study and explain the most important plots.

The drafter should begin from a copied example instead of inventing a schema. The runner uses the installed CLI. Reviewer and reporter skills use the same deterministic commands documented under Understanding Outputs.

Grouped Study Workflow

Use astraeus-study-manager when the question requires a parameter sweep or an explicit set of different RunSpecs. It can select the correct grouped mode, plan cases, choose bounded local workers, resume interrupted work, and rerun selected cases. Hand the completed bundle to astraeus-study-reporter for comparison.

$astraeus-study-manager Plan and run the copied altitude-trade-sweep example
with two workers. Report material progress and do not overwrite an existing study.

$astraeus-study-reporter Compare the completed altitude study, rank cases by
represented-area coverage, and call out any failed or degraded cases.

Campaign Workflow

Campaign skills operate the supported constellation-deployment surface. The standalone mission-planning workflow remains internal maintainer tooling.

$astraeus-campaign-composer Adapt the deployment-campaign example for
three planes with two satellites per plane and validate the request.

$astraeus-campaign-runner Run my-deployment-campaign.yaml and summarize the
closure status, selected strategy, result path, and any residual constraints.

$astraeus-campaign-analyzer Review the completed campaign result and explain
which target positions were achieved and what blocked the remainder.

Operational Boundaries

• Skills may create inputs, execute commands, and write derivative reviews or reports, so inspect the proposed paths before approving overwrites.
• A skill should not turn absent or degraded evidence into a successful claim.
• The Run Monitor, logs, and saved artifacts are the durable record for long-running work; chat updates provide supplementary progress context.
• astraeus-operations-manager is for an installed release and an environment that has already been qualified. Server provisioning, security, and initial qualification remain explicit administrator responsibilities.

Continue with Examples, the Console, or the Run Monitor.