Claude discovers only skills/<name>/SKILL.md, so 84 nested skills, including
all rule skills, were invisible. Flatten the categorized catalog for Claude,
keep references, and fail the build on duplicate names or a name that differs
from its directory.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Set CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS=5 and
CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH=1 so Claude matches the Codex agent
limits and subagents cannot start further subagents.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add a shell-specific permission for `mcporter call context7.*` so Claude can
query Context7 through MCPorter without a permission prompt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Search, Diff, Shell, and Tests are not valid Claude Code tools, so every
generated subagent silently launched with a broken or empty tool set. Map
each role to the tools its own prompt actually needs, and use a __SHELL__
placeholder so build resolves it to the target package's own shell tool
instead of granting both Bash and PowerShell unconditionally.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The Stop hook only runs a ~500ms visual flash; 20 seconds was far more
than any realistic process-start latency needs, including the one-time
native shim compile on Windows.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Adds adapters/claude/rule-skills.tsv, mapping each technology- or
situation-specific rule to a Claude skill name and trigger description.
Replaces the hardcoded rule-loading block in the shared instructions
template with a __RULE_LOADING__ placeholder so each client's build step
can substitute its own loading mechanism without duplicating the shared
orchestration text.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Julian lechnerandClaude Opus 5.5
fraujulian