feat: split ailss agent skill architecture and installer workflow#73
feat: split ailss agent skill architecture and installer workflow#73maehwasoo wants to merge 5 commits into
Conversation
…plit # Conflicts: # docs/ops/codex-skills/prometheus-agent/SKILL.md # docs/standards/vault/frontmatter-schema.md # docs/standards/vault/typed-links.md # packages/mcp/src/lib/ailssNoteTemplate.ts # packages/obsidian-plugin/src/settings.ts
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f6b990775e
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| - `typed-links.md` (index) | ||
| - `typed-links-relation-catalog.md` (relation semantics) | ||
| - `typed-links-decision-tree.md` (selection rules) | ||
| - `typed-links-governance.md` (SoT + change process) |
There was a problem hiding this comment.
Point ontology references to actual vault doc paths
The prompt installer now ships only prompt-operational.md (packages/obsidian-plugin/src/utils/promptTemplates.ts), but these references are bare filenames that do not match the repository layout where the canonical docs live under docs/standards/vault/. After installing the prompt at the vault root, assistants are directed to paths that typically do not exist, which breaks the intended follow-up guidance for typed-link decisions and governance; these should be written as repository-relative paths (for example, docs/standards/vault/...).
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 454f686c75
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| 1. Pull candidates with `get_context`. | ||
| 2. Verify exact source/target wording via `resolve_note` + `read_note`. | ||
| 3. Inspect graph context with `expand_typed_links_outgoing` and optional incoming checks. | ||
| 4. Update frontmatter with `improve_frontmatter` or `edit_note`. |
There was a problem hiding this comment.
Gate ontology skill write steps behind write-tool availability
This workflow instructs users to apply improve_frontmatter/edit_note unconditionally, but MCP write tools are disabled by default in AILSS deployments, so following the skill in the default setup will trigger deterministic tool-call failures instead of a usable read-only flow. The previous core skill explicitly marked write tools as conditional; the new modular skills should do the same (or provide a read-only fallback path) to avoid breaking default Codex sessions.
Useful? React with 👍 / 👎.
What
ailss-agent+ailss-agent-ontology+ailss-agent-curator+ailss-agent-maintenance.ailss-prometheus-agentas a one-release compatibility shim and updated Codex-skill docs references.Why
~/.codex/skillswithout manual copy-only workflow.How
docs/ops/codex-skills/*and updated legacy shim content.typed-links-relation-catalog.md,typed-links-decision-tree.md,typed-links-governance.md) and repurposedtyped-links.mdas an index.prompt-operational.mdand switched plugin prompt bundling to this concise prompt source.settings.ts,main.ts,registerCommands.ts, and newcodexSkillInstaller.ts.packages/mcp/test/docs.mcpToolingConsistency.test.tsto assert typed-link key parity and updatedailssNoteTemplate.tsto useAILSS_TYPED_LINK_KEYSdirectly.pnpm format:check,pnpm lint,pnpm typecheck,pnpm typecheck:repo, andpnpm test.