Skip to content

feat(skill): enhance agent skill with save rules, annotation conventions, and pipeline integration#14

Merged
gdiab merged 3 commits intomainfrom
feat/enhanced-agent-skill
Feb 24, 2026
Merged

feat(skill): enhance agent skill with save rules, annotation conventions, and pipeline integration#14
gdiab merged 3 commits intomainfrom
feat/enhanced-agent-skill

Conversation

@thoth-bot-ai
Copy link
Collaborator

What

Enhances the bundled agent skill (skills/linkledger-cli-agent/SKILL.md) to make it more useful for sub-agents and multi-agent workflows.

Changes

  • Setup section: documents wrapper path and DB config so agents don't need to figure out the environment
  • When to save: clear rules — save every cited source, every human-shared link, newsletter items; skip throwaway searches
  • Annotation conventions: actor naming (agent:<name>), confidence scale guidelines (0.5–1.0 with use-case table), highlight vs lowlight guidance, when to pin
  • Content pipeline integration: step-by-step for saving sources during content packet/blog research, linking item IDs to content-board cards

Why

The existing skill covers the how (command syntax) but not the when or why. Sub-agents doing research tasks need opinionated guidance on save discipline, annotation quality, and how LinkLedger fits into a broader content pipeline.

Closes #13 partially (documents the wrapper workaround in the setup section).

…tions, and pipeline integration

- Add 'When to save' section with clear rules for agents
- Add setup section with wrapper path and DB config
- Add annotation conventions: actor names, confidence guidelines, highlight vs lowlight
- Add content pipeline integration workflow (save → ingest → annotate → link to content-board)
- Clarify that linkledger wrapper works from any directory
@gdiab
Copy link
Owner

gdiab commented Feb 24, 2026

Review findings from my pass:

  1. Setup guidance is too environment-specific. skills/linkledger-cli-agent/SKILL.md currently hardcodes /opt/homebrew/bin/linkledger and says it can be called from anywhere. That won’t hold for non-Homebrew/non-macOS setups. Recommend making this conditional and keeping repo-local invocation as default fallback.

  2. No env var needed conflicts with repo behavior/docs. The code defaults DB path from current working directory (src/db/database.ts) and docs explicitly support LINKLEDGER_DB_PATH override (README.md). Current wording could send agents to the wrong DB unless wrapper assumptions are true.

Everything else in the PR (when-to-save, annotation conventions, pipeline integration) is a solid improvement.

- Remove hardcoded /opt/homebrew path assumption
- Document both wrapper and repo-local invocation modes
- Clarify LINKLEDGER_DB_PATH behavior and cwd default
@thoth-bot-ai
Copy link
Collaborator Author

Good catches — both fixed in 2504a1e:

  1. Removed hardcoded /opt/homebrew path. Setup now documents two modes: wrapper-based (if installed) and repo-local invocation as the default fallback.

  2. Clarified DB path behavior. Explicitly documents the cwd default, when to set LINKLEDGER_DB_PATH, and the risk of writing to unintended locations without it.

Copy link
Owner

@gdiab gdiab left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Re-review complete. The previously reported issues are resolved:

  • Setup guidance is now portable (wrapper is conditional, repo-local fallback documented).
  • DB path behavior is now aligned with project defaults and LINKLEDGER_DB_PATH override guidance.

No additional blocking findings from this pass.

@gdiab gdiab merged commit ee85a5e into main Feb 24, 2026
3 checks passed
@gdiab gdiab deleted the feat/enhanced-agent-skill branch February 24, 2026 23:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Bug: migrations directory resolved relative to cwd, not package root

2 participants