Oneshotprompt CLI

Use this skill when working inside the OneShotPrompt repository and the task involves the product CLI, scripts, docs, tests, or examples that exercise that CLI.

Primary commands:

1. dotnet restore OneShotPrompt.slnx
2. dotnet build OneShotPrompt.slnx
3. dotnet test tests/OneShotPrompt.Tests/OneShotPrompt.Tests.csproj
4. dotnet run --project src/OneShotPrompt.Console -- validate --config <path>
5. dotnet run --project src/OneShotPrompt.Console -- jobs --config <path>
6. dotnet run --project src/OneShotPrompt.Console -- run --config <path>
7. dotnet run --project src/OneShotPrompt.Console -- run --config <path> --job <name>
8. dotnet run --project src/OneShotPrompt.Console -- listen --config <path> --job <name>
9. dotnet run --project src/OneShotPrompt.Console -- interactive

Workflow:

1. Prefer explicit CLI commands over the no-argument launch path when writing automation, docs, tests, or examples.
2. Treat the config file directory as the runtime root for sibling skills/, logs/, and .oneshotprompt/memory/ content.
3. Validate a config before assuming a job can be listed or run.
4. List jobs before referencing a job name that was not already established.
5. When documenting or testing single-job execution, use run --config <path> --job <name>.
6. When documenting the WhatsApp personal channel flow, use listen --config <path> --job <name> if the intent is continuous event-driven replies rather than a single pass.
7. When describing interactive behavior, distinguish between explicit interactive and the no-argument behavior that depends on whether the terminal is interactive.
8. For build or test validation in this repo, use the solution and test commands above rather than ad hoc project combinations unless the task is intentionally scoped to one project.
9. If jobs or config examples mention execution behavior, account for both Workflow: "single-agent" and Workflow: "corporate-planning" where relevant.
10. If you change CLI behavior, command examples, scheduling flows, or bundled skills, update README.md and the relevant docs/ pages in the same change.

Guardrails:

• Do not document or script unsupported subcommands or flags.
• Do not treat --job as valid for validate or jobs.
• Do not describe listen as a general scheduler; it is a targeted bridge-driven listener mode.
• Do not describe logs or memory as repo-root relative unless the active config file is at the repo root.
• Do not describe every run as a single execution agent when a job can opt into the corporate-planning workflow.
• Do not rely on shell-specific syntax in examples unless the task is explicitly shell-specific.