cronekorkn/dotfiles
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

docs(agents): Subagents-Sektion überarbeitet — Delegations-Default, fork, SendMessage statt task_id

Browse source 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017wcGPTAnGpH7wFGXCGQjtw
This commit is contained in:
mwiegand 2026-08-30 14:27:23 +02:00
parent c77790dae1
commit 8db8e6a3c8
No known key found for this signature in database
1 changed files with 33 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

42
agents/AGENTS-claude.md
View file

@ -11,12 +11,30 @@
- Two shortcuts past onyx: Thicket MCP for people lookups (role, team,
email — its other sections are thin), Confluence CQL for precise
lookups with known space/title.
- Same-topic follow-ups: use `onyx answer "..." --session <last session_id>`.
## Subagents
Spawn subagents to (a) isolate context-heavy work from the main thread, or (b) parallelize independent subtasks. Before spawning, ask: *could `Bash`/`Grep`/`Read` do this directly?* Subagents have overhead — skip the middleman for mechanical lookups.
**Default to delegating whenever the expected tool output exceeds the
conclusion the main thread actually needs** — multi-file research, log/data
analysis, doc review, broad searches, verification runs. The session context
is the scarce resource; a subagent burns its own context and returns only the
result. Work inline only for single-fact lookups where file and symbol are
already known — there, a subagent is pure overhead.
Use the `Agent` tool. Prefer specialized agent types (`Explore`, `code-reviewer`, `Plan`, etc.) over `general-purpose` when one fits.
Use the `Agent` tool. Prefer specialized agent types over `general-purpose`
when one fits:
- **`Explore`** — read-only search: *locates* code and facts (reads excerpts,
not whole files). It does not review or audit; quality judgments stay with
the parent or a dedicated reviewer.
- **`Plan`** — implementation strategy, architecture trade-offs.
- **`fork`** (`subagent_type: "fork"`) — inherits the full conversation
context, runs in the background, keeps its tool output out of the main
thread. The tool of choice for "do X with everything you already know, give
me only the outcome" — no need to re-explain state in the prompt.
- Agent types not listed as available in the session (e.g. a `code-reviewer`)
don't exist there — check the available-types list instead of guessing.
### Behavioral tier (MANDATORY — always set)
@ -28,15 +46,21 @@ Prefix every `Agent` prompt with one of:
### Model selection (override only when needed)
Resolution order: explicit `model` param → agent type frontmatter → parent inheritance.
Resolution order: explicit `model` param → agent type frontmatter → parent
inheritance. (`fork` always runs on the parent model; an override is ignored.)
1. **Don't override specialized agent types** (`Explore`, `code-reviewer`, `Plan`, etc.) — their frontmatter is tuned. Leave `model` off.
2. **For `general-purpose` under an Opus parent: set `model: sonnet`** — avoid paying Opus rates for routine subagent work.
3. **`model: opus`** on a subagent only when isolated context *and* reasoning beyond what the parent can easily do inline are both needed.
1. **Don't override specialized agent types** (`Explore`, `Plan`, etc.) — their frontmatter is tuned. Leave `model` off.
2. **For `general-purpose` under an expensive parent (Opus/Fable): set `model: sonnet`** — don't pay top rates for routine subagent work.
3. **A bigger model (`opus`)** on a subagent only when isolated context *and* reasoning beyond what the parent can easily do inline are both needed.
4. **`model: haiku`** only for mechanical tasks: bulk classification, format conversion, summarizing pre-filtered text. Code-semantic work → Sonnet.
### Context hygiene
- Prefer parallel `Explore` calls over sequential `general-purpose` sessions.
- Don't reuse `task_id` for fresh lookups; start new to keep prompts tiny.
- Summarize subagent findings in the main thread to keep the terminal clean.
- Prefer parallel `Explore` calls over sequential `general-purpose` sessions;
spawn independent agents in a single message so they run concurrently.
- Once delegated, don't duplicate: never run the same search yourself while
an agent is on it, and don't poll — results arrive as notifications.
- Follow-ups to an existing agent go via `SendMessage` (keeps its context);
a new `Agent` call always starts fresh — right for unrelated lookups.
- Summarize subagent findings in the main thread; the raw report is not
shown to the user, so relay what matters.