Skills Reference

The Tracer Spec Workflow ships seven skills. Each is a self-contained instruction file; this page catalogs every skill’s trigger, argument, and tool access. The SKILL.md files are the source of truth — this page mirrors their frontmatter.

Skill Purpose Argument Tools

Skill	Purpose	Argument	Tools
`tspec-explore`	Shape a change sketch without writing.	`<topic>`	unrestricted (read-only)
`tspec-propose`	Write a change, its requirements, and decisions atomically.	`<change-id or title>`	`antora-tracer`
`tspec-apply`	Implement a change’s tasks and advance state.	`<change-id>`	`antora-tracer`
`tspec-archive`	Validate, verify, and close a change.	`<change-id>`	`antora-tracer`
`tspec-validate`	Validate the traceability graph.	—	`antora-tracer`
`tspec-write-item`	Write one item of any role.	`<role> [<id>]`	`antora-tracer`
`tspec-track`	Bridge a change to its issue-tracker issue.	`←-link\|--create\|--import\|--status> <id>`	`antora-tracer`, `gh`, `curl`

tspec-explore

Shape a change sketch without writing.

<topic>

unrestricted (read-only)

tspec-propose

Write a change, its requirements, and decisions atomically.

<change-id or title>

antora-tracer

tspec-apply

Implement a change’s tasks and advance state.

<change-id>

antora-tracer

tspec-archive

Validate, verify, and close a change.

<change-id>

antora-tracer

tspec-validate

Validate the traceability graph.

—

antora-tracer

tspec-write-item

Write one item of any role.

<role> [<id>]

antora-tracer

tspec-track

Bridge a change to its issue-tracker issue.

←-link|--create|--import|--status> <id>

antora-tracer, gh, curl

tspec-explore

Enter explore mode — a thinking partner that produces a change sketch (rationale, outcome, scope) without writing durable content. Read-only by discipline: it may read files, search code, and run commands, but must not edit the repo or docs/modules/specs/pages/.

tspec-propose

Propose a change: write the change item, its requirements, and its decisions in one pass, plus the temporary task checklist at tspec/<change-id>/tasks.md. Generates every ID up front, resolves role guidance templates, writes each item as its own file with a matching include:: line, and validates before reporting. If tspec.yml declares a tracker, delegates issue linking to tspec-track.

tspec-apply

Implement an accepted change: work the task checklist, tick - [ ] to - [x], add verification evidence for each addressed requirement, and advance accepted → implemented → verified. Never modifies requirements in place — a changed meaning is a supersession.

tspec-archive

Close a verified change: validate, spawn an independent cold reviewer that verdicts each requirement SATISFIED/NOT, then delete tspec/<change-id>/ and advance verified → closed. Stops on any NOT-satisfied verdict.

tspec-validate

Run antora-tracer validate and report the summary plus any errors with file and line. Never changes state.

tspec-write-item

Write one item of any declared role (change, requirement, decision, test, or a declared growth role) and wire its relations directly. Generates the next ID from the role’s prefix and writes into the role’s page.

tspec-track

Bridge a change to its issue tracker, one-way.

Subcommand Behavior

Subcommand	Behavior
`--link <change-id>`	Record an existing tracker issue as `trackerId=` on the change.
`--create <change-id>`	Create a tracker issue for the change, then record its ID.
`--import <issue-id>`	Import a tracker-first issue as a seeded change.
`--status <change-id>`	Push the change’s lifecycle status to its linked issue.

--link <change-id>

Record an existing tracker issue as trackerId= on the change.

--create <change-id>

Create a tracker issue for the change, then record its ID.

--import <issue-id>

Import a tracker-first issue as a seeded change.

--status <change-id>

Push the change’s lifecycle status to its linked issue.

Configuration lives in tspec.yml under tracker. provider: none (the default) makes every subcommand a no-op. github drives the gh CLI; jira drives REST against the configured url and project. Tracker state is never read back into the graph; trackerId is a block-header attribute, not a relation.