Commit Graph

30 Commits

Author SHA1 Message Date
Levi Neely 36010d0143 prompts: add tools-lsp prompt, wire into all agents 2026-04-30 11:14:13 +02:00
Levi Neely c37a4db480 freeloader: fix hook wiring; remove || true from agent turnError hooks
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-28 17:27:34 +02:00
Levi Neely 15df9eeb1a freeloader: turnError-driven free model cycling with blacklist
- scripts/x/freeloader: hook-driven script; intercepts rate_limit and
  tool_unsupported on :free models; auto-blacklists on tool_unsupported;
  cycles to next non-blacklisted free model; exits 1 for unhandled errors
- agents: add turnError hook to all agent configs

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-28 17:17:56 +02:00
Levi Neely 2f9e392320 restore reasoning_think tool and prompt
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 19:07:13 +02:00
Levi Neely e4c76603ba remove reasoning_think tool and prompt
Backend now streams native thinking tokens via StreamEvent.Reasoning.
The no-op scratchpad tool is redundant.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 19:04:03 +02:00
Levi Neely 18a7840b3c agents: add SYSTEM_PROMPT to prompt array, move all prime calls from hooks 2026-04-26 15:10:22 +02:00
Levi Neely d333bf290a agents: extract prompts to markdown, assemble via executable prompt
Move inline JSON prompt strings to prompts/agent-*.md files, referenced
via $OLLIE/x/prime in the new executable prompt array. Move all
prime calls (tools-*, beads) from hooks.agentSpawn into prompt.

Rename greybeard to sysarchitect.
2026-04-26 15:00:46 +02:00
Levi Neely b5a0765fc1 tools/agents: skip blank task_add args; remind task_check when plan has open tasks
preTurn now has two branches: no plan → remind task_add; open tasks → remind
task_check. Blank/whitespace-only arguments to task_add are silently skipped.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 13:26:03 +02:00
Levi Neely 70f9aed14e agents: restore preTurn task_add reminder
Removing it caused the model to fall back to reasoning_think for planning.
The preTurn reminder is necessary to override that behavior.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 13:13:48 +02:00
Levi Neely f4b6cfa4f9 agents: revert preTurn reminder; tools-task prime is sufficient
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 13:12:23 +02:00
Levi Neely 2d20f95b93 agents: name task_add explicitly in preTurn reminder
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 13:07:59 +02:00
Levi Neely 68246cd527 agents: preTurn reminder to plan before acting when plan is empty
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 13:04:46 +02:00
Levi Neely 84e7b29015 agents: inline bd prime conditional; simplify task dispatcher
Replace 'task prime' hook with inline .beads walk in all agent configs —
no wrapper script needed. Drop 'prime' subcommand from scripts/x/task.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 12:31:34 +02:00
Levi Neely b7f6b08e6a task: session plan checklist; replace todo.sh fallback
- tools/task_add, task_check, task_clear: manage s/{sid}/plan markdown checklist
- prompts/tools-task.md: agent command reference, injected at spawn via prime tools-task
- agents: add 'prime tools-task' to agentSpawn in all configs
- scripts/x/task: show_plan() on 'task list'; drop todo.sh fallback (wrong tool)
- prompts/tools-todosh.md: removed (no longer needed)
- prompts/tools-subagent.md: add coordination model documentation
- core: update submodule ref (s/{sid}/plan 9p endpoint)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 12:30:14 +02:00
Levi Neely 1a01002f7b planning: observe step, task dispatcher, tool parallel annotations, graceful degradation
- agents: agentSpawn uses 'task prime', postTurn uses 'task list' (observe step)
- scripts/x/task: dispatcher routes to bd (beads) or todo.sh based on availability
- prompts/tools-todosh.md: prime context for todo.sh fallback
- tools: add '# ollie:parallel read' annotation to file_read, file_glob, file_grep,
  file_grep, memory_recall, reasoning_think, denote_view
- core: update submodule ref

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-26 09:00:54 +02:00
Levi Neely c170917ff2 agents: revert postTurn bd list (premature)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-25 23:56:31 +02:00
Levi Neely c0a4fd1525 agents: observe step via postTurn bd list
All agents now run `bd list --status open` after each turn. Output is
injected into session history (not stdout) so the model is aware of open
tasks on the next turn without triggering autonomous continuation.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-25 23:53:11 +02:00
Levi Neely df4d8a3a19 agents: move tool primes to agentSpawn; update core submodule
Tool primes (tools-file, tools-reasoning, tools-memory, tools-subagent,
tools-elevate) moved from preTurn to agentSpawn across all agent configs.
Context is now injected once at session refresh points (start, post-clear,
post-compaction) rather than repeated on every turn.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-25 19:38:27 +02:00
Levi Neely 2f8d18dc51 elevate: auto-start adapter via agentSpawn hook
Add x/elevate-adapter-start: checks if the adapter is already listening
(connect test), starts it in the background if not. Exits 0 always so it
never blocks session startup. Wire it into agentSpawn in all agent configs
and install it via mkfile.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-25 11:54:29 +02:00
Levi Neely 7e62336c61 prompts: consolidate sys-* into SYSTEM_PROMPT.md; load via Go, not hooks
sys-base, sys-ollie, sys-skills merged into a single SYSTEM_PROMPT.md.
Core now loads and envsubst-expands it at startup, prepending it to the
agentPrompt to form the system message. Removes the three preTurn hook
injections from all agent configs.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-21 09:14:14 +02:00
Levi Neely c8d026a3da agents: coding prompt add good output criteria 2026-04-20 22:40:01 +02:00
Levi Neely c969805813 agents: move sys-* prompts to preTurn 2026-04-20 22:35:12 +02:00
Levi Neely 61fc50a702 prompts: rename p/ files; drop OLLIE_PLAN_PATH and pl/ references
Rename prompt files to semantic names (no numerical prefix):
  00_base           => sys-base
  01_ollie          => sys-ollie
  02_skills         => sys-skills
  03_tool-files     => tools-file
  04_tool-reasoning => tools-reasoning
  05_tool-memory    => tools-memory
  07_tool-subagents => tools-subagent
  08_tool-elevate   => tools-elevate

Update agent configs and all documentation references. Remove
OLLIE_PLAN_PATH from env.sample and pl/ mention from OBSERVATIONS.md.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-19 20:09:50 +02:00
Levi Neely 1ee187b086 agents: rename userPromptSubmit→preTurn, stop→postTurn; bump core
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-19 15:47:51 +02:00
Levi Neely aa0d01e60f agents: use $OLLIE/x/prime absolute path in agentSpawn hooks
prime may not be in PATH at hook execution time; use full path to avoid
the race.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-19 12:13:28 +02:00
Levi Neely 8d870001bc prompts: migrate to hook-based priming via prime script
Remove Go template syntax from all prompt files (converted to envsubst
variables in prior commits). Agent agentSpawn hooks now call prime for
each prompt file explicitly, replacing the automatic Go-based directory
scan. coding agent gets 08_tool-elevate; default does not.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-19 12:10:26 +02:00
Levi Neely 7c10a127c4 agents: add bd prime to coding agentSpawn; add README; remove test-bd
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-19 11:57:55 +02:00
Levi Neely 51a20195bf agents: convert hook values to array format
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-19 11:54:56 +02:00
Levi Neely c5d87567ce prompts: restructure system prompt; replace task tools with plan_* lifecycle
- Rename and reorder prompt files: skills bumped to 02, tool prompts
  prefixed tool-*, execute_code prompt removed (content absorbed into schema)
- Rewrite all tool prompts with explicit Trigger/Calling convention/Returns/Constraints structure
- Trim 03_reasoning.md to essentials; add "returns nothing" clarification
- Restructure task planning: replace task_plan/task_complete with explicit
  plan_create/plan_start/plan_check/plan_finish lifecycle tools
  - plan_finish enforces hard constraint: fails if any - [ ] tasks remain
- Remove no-shell-redirect rule from 00_base (consolidated into 03_tool-files)
- Add t/idx discovery and parallel schema advice to execute_code JSON schema
- Remove dead execute_pipe/execute_tool trustedTools references from agent configs
- Make keyword greps case-insensitive (-i flag)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-18 14:48:10 +02:00
Levi Neely c62a599f71 prompts/agents: stronger no-narration rule; split default/coding agents
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-18 14:05:46 +02:00