Replace Claude-based link building with direct subprocess calls to
Big-Link-Man CLI. The build stage now runs ingest-cora + generate-batch
deterministically without LLM involvement.
- New clickup_runner/blm.py: BLM subprocess runner (ported from old linkbuilding.py)
- New handler="blm" in skill_map for Link Building build stage
- Add BLMConfig to config.py (blm_dir, cora_inbox, timeout)
- Add task-specific fields (Keyword, CLIFlags, etc.) to build_prompt
- Strip YAML frontmatter from skill files in read_skill_file
- Skip orphaned AutoCora results with no state DB entry
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Previously, results from prior runs (or the old CheddahBot system) would
be processed without context, causing "unknown" keyword in comments and
cascading errors. Now they are archived silently.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix --tier-1 -> --tier1-count in create_task_set.py and create-task-set.md skill
- Set initial Stage field when creating tasks (run_cora for most, draft for PR)
- Add Quick Reference section to clickup_runner README with stage tables
- Add Stage field docs to clickup-task-creation.md
- Remove stale xlsx_urls tests from test_claude_runner.py
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The ClickUp dropdown is named "Client", not "Customer". Updated field
lookups, prompt labels, README docs, and all tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Submits Cora SEO jobs as JSON files to the NAS queue, polls for .result
files each cycle, and updates ClickUp on success/failure. Async design:
submission and result processing happen on separate poll cycles so the
runner never blocks. Full README rewrite with troubleshooting section.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implements _dispatch_claude: reads skill files, builds prompts with task
context, runs claude -p as subprocess, uploads output files to ClickUp,
copies to NAS, advances stage/status, and posts structured error comments
on failure. Includes ntfy.sh notifications and generous max_turns defaults.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New headless automation runner that replaces CheddahBot's multi-agent
system with a simpler design: poll ClickUp for tasks with "Delegate to
Claude" checkbox checked + due date <= today, route by task type + stage
through a skill map, and dispatch to Claude Code headless or AutoCora.
Module structure:
- config.py: env vars > YAML > defaults configuration
- clickup_client.py: ClickUp API client (adapted from cheddahbot/clickup.py)
with checkbox, stage dropdown, and attachment operations
- skill_map.py: task_type + stage -> SkillRoute routing dictionary
(Content Creation, On Page Optimization, Press Release, Link Building)
- state.py: minimal SQLite KV store + run log
- __main__.py: poll loop with graceful shutdown
- README.md: setup, config reference, skill map docs
Dispatch functions are stubs (Phase 2: Claude runner, Phase 3: AutoCora).
Includes system-design-spec.md that drives this rewrite.
68 new tests, all passing. Existing 347 tests unaffected.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
PeninsulaInd