0030 — Skills of the harness's own, the same on every adapter
- Status: accepted (journal#971; built in journal#987) · Asked by: the project owner, 2026-11-11 (journal#888; and 'скилы стоит сделать свои для переносимости между адаптерами', journal#905). The agent drafted the design overnight.
- Specs: CON-135, CON-145 (the delegates' own plugins and skills stay theirs)
Why
A skill written once should work whatever runs the conversation: the harness's own loop, Claude Code or Codex. Claude Code and Codex each have skills of their own, and those keep working inside them (CON-145). The harness's skills are the portable layer above them.
Decision (draft)
Where skills live
They are laid out as Claude Code lays out its own, with .harness in place of .claude:
<project>/.harness/skills/<folder>/SKILL.md the project's
~/.harness/skills/<folder>/SKILL.md the user's, global (MULTITOOL_HOME when set)
SKILL.md starts with YAML frontmatter with name and description, and the body is the instructions. Files beside it (scripts, templates) are the skill's own, read by path.
- External contract, verify before building: Anthropic's Agent Skills format.
- Frontmatter fields
name(lowercase, digits, hyphens) anddescriptionare required;allowed-toolsis optional. - Verify: docs.claude.com, Agent Skills; record it in an
### External contractsentry with the doc's URL.
- Frontmatter fields
- The same names on both levels: the project's wins (journal#888).
- A skill folder of Claude Code's (
.claude/skills) is not read by the harness: it is Claude Code's own, and stays so (CON-145).
How an agent uses them
Two tools are served by the harness's MCP server to every adapter. The native loop gets them as tools; Claude Code and Codex get them through the harness server they are already given.
skill_list: the skills here, each with its name, description, and whether it is the project's or the user's. Only descriptions, so a long list costs little.skill_load(name): the skill's body, and the paths of its files.
The specification of the agent workflow gains a row: when a task matches a skill's description, skill_load it first.
On the screen
/skillslists them.- A skill loaded in a turn shows as a card,
SKILL name, folded.
Open questions for the owner
- Should the harness also offer the delegates' own skills (
~/.claude/skills) to the native loop and Codex, read-only? The draft says no; each program keeps its own, and only.harness/skillsis portable. - Should a skill be able to declare which adapters it is for? The draft says no, a skill is for all.