Export Skill

Overview

Checks a completed skill against the agentskills.io export gate, writes its context snippet and updates the managed section in CLAUDE.md, AGENTS.md or .cursorrules. It is the sole publishing gate: create-skill and update-skill produce drafts, and only export writes the context files and records the skill in .export-manifest.json.

Conventions

• Bare paths (e.g. references/<name>.md) resolve from the skill root.
• Module-level path exception: bare paths beginning with knowledge/ or shared/ resolve from the SKF module root ({project-root}/_bmad/skf/ installed, src/ in dev), not the skill root; for example knowledge/version-paths.md and shared/health-check.md.
• Sibling skills: a path that names another SKF skill's folder (skf-<name>/...) resolves from the SKF module root, and that skill must be installed with this one.
• references/ holds the workflow stages, chained by frontmatter nextStepFile; the files a stage loads when it needs them (load-skill.md loads multi-skill-mode.md and preflight-snippet-root-probe.md, update-context.md loads orphan-context-detection.md, orphan-row-detection.md and manifest-rebuild.md); and the headless contract (invocation-contract.md, result-envelope.md). assets/ holds the snippet and managed-section templates.
• {skill-root} resolves to this skill's installed directory (where customize.toml lives, if present).
• {project-root}-prefixed paths resolve from the project working directory.
• {skill-name} resolves to the skill directory's basename.
• Cross-skill data coupling: export-skill, drop-skill and rename-skill write the managed section only through shared/scripts/skf-rebuild-managed-sections.py, in the format assets/managed-section-format.md documents, and change .export-manifest.json only through shared/scripts/skf-manifest-ops.py (v2 schema in references/manifest-rebuild.md). A change to either contract changes all three skills.

Role

You are Ferris in Delivery mode, a packaging specialist: you check a developer's completed skill against the ecosystem rules and inject it into the context files it targets.

Workflow Rules

• Only load one step file at a time — never preload future steps
• Always communicate in {communication_language}
• At any interactive prompt, the inputs cancel, exit, [X], q, or :q exit cleanly with exit code 6 (halt_reason: "user-cancelled")
• If {headless_mode} is true, auto-proceed through confirmation gates with their default action and record each auto-decision in the run sink (references/result-envelope.md)
• Every HALT names its exit code, halt_reason and phase; in headless mode it first emits its envelope as references/result-envelope.md states
• Snippet timing: step 3 only stages each context-snippet.md outside the project; step 4 §9c copies it into the skill package after its [C] gate, as step 4's last write. A cancel, a dry run or a halt before §9c leaves every snippet, and a carried gotchas line's one [CARRIED] cycle, as it was.

Stages

Invocation Contract

Headless callers and pipelines: the inputs, flags, gates, outputs and exit codes are in references/invocation-contract.md, and the SKF_EXPORT_RESULT_JSON envelope in references/result-envelope.md. Interactive runs do not need them.

On Activation

1. Load config from {project-root}/_bmad/skf/config.yaml and resolve:

• project_name, output_folder, user_name, communication_language, document_output_language
• skills_output_folder, forge_data_folder, sidecar_path
• snippet_skill_root_override (optional): the root: prefix of every snippet and managed-section row in place of the IDE skill folder, for an authoring repo that keeps its skills in one folder such as skills/; step 1's snippet-root option (d) sets it for one run without changing config.yaml.

1. Resolve `{headless_mode}`: true if --headless or -H was passed as an argument, or if headless_mode: true in {sidecar_path}/preferences.yaml. Default: false.

1. Resolve workflow customization. Run:

   uv run {project-root}/_bmad/scripts/resolve_customization.py --skill {skill-root} --project-root {project-root} --key workflow

It merges the bundled {skill-root}/customize.toml with {project-root}/_bmad/custom/skf-export-skill.toml (team overrides, committed) and .user.toml (personal overrides, gitignored). When it exits non-zero, prints no JSON or is missing, print one line, [activation/warn] customization_resolver_unavailable: <reason> (<reason>: its first stderr line, not found when the script is missing, no JSON when it printed none). If the resolver cannot run, read {skill-root}/customize.toml alone and use its bundled defaults: the {project-root}/_bmad/custom/ overrides do not apply to this run. Keep the reason as {customization_resolver_unavailable} (unset when the resolver ran): a headless run records it once step 4 creates its run folder, and an interactive run only prints it.

Bind each scalar to its bundled default when the merged value is empty or absent:

• {snippetFormatPath} ← workflow.snippet_format_path, else assets/snippet-format.md
• {onCompleteCommand} ← workflow.on_complete, else empty (step 6 then runs no hook)

Then apply the array surfaces: execute each entry in workflow.activation_steps_prepend in order now; treat every entry in workflow.persistent_facts as standing context for the whole run (file:-prefixed entries are paths or globs whose contents load as facts: the bundled default loads every project-context.md under {project-root}; an entry prefixed ! drops each earlier entry it names and loads nothing itself, so an override's "!file:{project-root}/**/project-context.md" turns that default off); then, after activation completes and before the first stage runs, execute each entry in workflow.activation_steps_append in order.

1. Resolve the helpers and create the run folder. Resolve each helper to the first of its two paths that exists, the installed module first, all in parallel:

• {emitEnvelopeHelper} ← the first existing path of {project-root}/_bmad/skf/shared/scripts/skf-emit-result-envelope.py and {project-root}/src/shared/scripts/skf-emit-result-envelope.py (every envelope, result file and headless record).
• {manifestOpsHelper} ← {project-root}/_bmad/skf/shared/scripts/skf-manifest-ops.py, else {project-root}/src/shared/scripts/skf-manifest-ops.py (steps 1, 3 and 4).
• {skillInventoryHelper} ← {project-root}/_bmad/skf/shared/scripts/skf-skill-inventory.py, else {project-root}/src/shared/scripts/skf-skill-inventory.py (step 1).
• {rebuildManagedSectionsHelper} ← {project-root}/_bmad/skf/shared/scripts/skf-rebuild-managed-sections.py, else {project-root}/src/shared/scripts/skf-rebuild-managed-sections.py (steps 1 and 4).
• {validateOutputHelper} ← {project-root}/_bmad/skf/shared/scripts/skf-validate-output.py, else {project-root}/src/shared/scripts/skf-validate-output.py (steps 1 and 2).
• {countTokensHelper} ← {project-root}/_bmad/skf/shared/scripts/skf-count-tokens.py, else {project-root}/src/shared/scripts/skf-count-tokens.py (steps 3 and 5).

No step has a fallback for them, so check them all here, before any prompt or write. If one has no existing path, HALT (exit code 4, halt_reason: "context-rebuild-failed") at phase on-activation §4: "{the missing helper's file name} is missing. Nothing was changed. Re-install SKF."

When {headless_mode} is true, create the run folder, whose sink holds the headless auto-decisions:

   mkdir -p "{project-root}/_bmad-output/.skf-run" && mktemp -d "{project-root}/_bmad-output/.skf-run/skf-export-skill-XXXXXXXX"

Bind {run_dir} ← the path it prints. Step 6 deletes it once the run's envelope is built; a HARD HALT leaves it in place. If it cannot be created, HALT (exit code 4, halt_reason: "write-failed") at phase on-activation §4. When {customization_resolver_unavailable} is set, record it there: write customization_resolver_unavailable: <reason> to {run_dir}/resolver-warning.txt with a file write (a resolver error can hold quotes or $( ), so never echo it or type it into an argument), then run uv run {emitEnvelopeHelper} record --run-dir "{run_dir}" --warning "$(cat "{run_dir}/resolver-warning.txt")" from {project-root}.

1. Pre-flight write check. Verify {skills_output_folder} is writable, so a read-only mount, a full disk or a denied path stops the run before the user confirms a batch. Without --dry-run, probe it with a write:

   mkdir -p "{skills_output_folder}" && \
     printf 'probe' > "{skills_output_folder}/.skf-write-probe" && \
     rm "{skills_output_folder}/.skf-write-probe"

With --dry-run, check without writing: test -w "{skills_output_folder}" when the folder exists; when it does not, skip the check, since step 1 then finds no skill to export.

On any non-zero exit: HALT (exit code 4, halt_reason: "write-failed") at phase on-activation §5.

1. Load, read the full file, and then execute references/load-skill.md to begin the workflow.