deep-execution

1# Agent-Enhanced Council Execution
2 
3Run one Workflow of parallel Claude analyst agents for deeper analysis. Each analyst
4queries its provider, evaluates response quality, can ask follow-up questions, and
5returns a structured analysis that the workflow enforces against the schema.
6 
7## Step 1: Resolve Providers and Write the Questions
8 
9One shell call resolves what the workflow needs — each provider's model, its
10role, and the question file it will send — with the same helpers the standard
11flow uses. Paste the final question into the heredoc verbatim: the quoted
12marker means the shell does NOT interpret quotes, backticks or `$()` in it.
13The question is emitted once; each provider's role-injected variant is built
14in shell, not pasted again.
15 
16```bash
17source "${CLAUDE_PLUGIN_ROOT}/scripts/lib/providers.sh"
18source "${CLAUDE_PLUGIN_ROOT}/scripts/lib/roles.sh"
19PROVIDERS=(<the selected providers, space separated>)
20ROLES="<the --roles value, or empty>"
21RUN=$(date +%s)                                  # names this run's files; Step 6 reuses it
22Q="$PWD/.claude/council-cache/.agents-$RUN"
23mkdir -p "$PWD/.claude/council-cache"
24# Self-ignoring, as cache.sh and run-council.sh keep it: these files carry the
25# question and any --file contents, and must never land in a commit.
26[[ -f "$PWD/.claude/council-cache/.gitignore" ]] || printf '*\n' > "$PWD/.claude/council-cache/.gitignore"
27cat > "$Q.txt" <<'COUNCIL_Q_EOF'
28<the question, verbatim>
29COUNCIL_Q_EOF
30…