Personal agent assets managed with dotagents. The skill source repo is dcramer/agents; dotagents documentation lives at dotagents.sentry.dev.

Install skills into your user-level agent configuration:

npx @sentry/dotagents init --user
npx @sentry/dotagents add --user dcramer/agents garfield
npx @sentry/dotagents add --user dcramer/agents garfield-codify
npx @sentry/dotagents install --user

dotagents init --user creates a user-level agents.toml. dotagents add --user records this repo as the skill source, and dotagents install --user installs declared skills into the user agent skill directory.

Example agents.toml entry:

version = 1
agents = ["codex", "claude", "cursor"]

[[skills]]
name = "garfield"
source = "dcramer/agents"

[[skills]]
name = "garfield-codify"
source = "dcramer/agents"

You can also install this repo's skills with the skills CLI:

npx skills add dcramer/agents --skill garfield
npx skills add dcramer/agents --skill garfield-codify

Install every skill in this repo:

npx skills add dcramer/agents --all

• skills/ contains reusable agent skills.
• skills/garfield/ runs an incremental implementation review-fix-verify loop with evidence-labeled concerns and optional independent verification.
• skills/garfield-codify/ analyzes Garfield findings and plans deterministic lint, type, CI, schema, generated-artifact, or pre-commit checks.

The implementation loop is named garfield instead of a generic name like iterate so it stays unique across personal and shared skill registries and avoids conflicts with similarly named workflow skills.

Each skill lives in its own directory and starts with SKILL.md. New skills should include:

• SKILL.md for runtime instructions.
• SPEC.md for the maintenance contract when the skill intent or scope is non-trivial.
• SOURCES.md for provenance, decisions, gaps, and changelog entries.
• agents/openai.yaml for optional OpenAI UI metadata, such as display name, short description, and default prompt.