Commands¶

ossature init¶

Create a new project.

ossature init myproject

Creates a directory with ossature.toml and a specs/ folder. Use . to initialize the current directory.

ossature new¶

Create a new spec file from a template.

ossature new my-feature              # creates specs/my-feature.smd
ossature new my-feature -t amd       # creates specs/my-feature.amd
ossature new my-feature -i           # interactive mode

ossature validate¶

Parse all .smd and .amd files and check for structural issues. No LLM calls.

ossature validate

Checks that every @depends target exists, every @spec reference in AMDs resolves to a real SMD, there are no duplicate component names within a spec, and there are no cycles in the dependency graph.

If a spec has high requirement complexity, validate prints a warning. Validation still passes, but the spec may cause problems during plan generation. Consider splitting it into smaller specs linked with @depends.

ossature audit¶

The most complex command. Sends specs to the LLM for review, generates context files, and produces a build plan.

ossature audit

What it does, in order:

1. Validates everything (same checks as validate)
2. Builds the spec dependency graph, writes .ossature/graph.toml
3. Computes checksums of all source files, compares to saved manifest. If nothing changed, skips re-audit
4. Audits each changed spec with the LLM, looking for ambiguity, contradictions, gaps, and feasibility issues
5. Auto-fixes errors (up to 3 cycles per spec), re-auditing after each fix
6. Runs a cross-spec audit if there are multiple specs, checking for interface mismatches
7. Writes the audit report to .ossature/audit-report.md
8. Regenerates the project brief and per-spec briefs whose inputs changed since the last audit
9. Extracts or infers interface signatures for each spec
10. Generates the build plan, writes .ossature/plan.toml
11. Exits with code 1 if any audit errors remain unresolved

By default, the audit runs non-interactively: it auto-fixes errors without prompting, prints a consolidated findings table at the end, and exits with code 1 if errors remain.

When only some specs changed, audit runs incrementally: only the changed specs are re-planned. The planner receives a diff of what changed and the previous task plan, so it can keep unaffected tasks stable. Tasks in the changed spec that produce the same output files as before carry over their build status. Tasks for unchanged specs are preserved with their existing hashes. Stale output files from tasks that no longer exist in the new plan are automatically removed.

Flags:

--interactive and --no-fix are mutually exclusive.

ossature audit --replan         # force full plan regeneration
ossature audit -i               # interactive mode with prompts
ossature audit --no-fix         # just show findings, don't fix
ossature audit --errors-ok      # don't fail on remaining errors

ossature build¶

Execute the build plan, generating code task by task.

ossature build

Flags:

--step and --auto are mutually exclusive. --skip-failures requires --auto.

In the default mode (no flags), the build continues silently on success and pauses on failure with a prompt: retry, skip, or quit.

ossature retry¶

Re-run failed or specific tasks.

ossature retry                # re-run all failed tasks
ossature retry --from 007     # redo everything from task 007 onwards
ossature retry --only 005     # re-run task 005 and all its dependents

--from and --only are mutually exclusive.

With --only, Ossature walks the dependency graph to find every task that transitively depends on the specified task and marks those as pending too.

ossature status¶

Show current build progress.

ossature status

Shows how many tasks are done, pending, failed, or skipped.

ossature clean¶

Remove the .ossature/ directory. Full reset.

ossature clean

Global Options¶