SessionStart

node "${CLAUDE_PLUGIN_ROOT}/scripts/hooks/inject-skill-context.js"

node "${CLAUDE_PLUGIN_ROOT}/scripts/hooks/init-data-dir.js"

node "${CLAUDE_PLUGIN_ROOT}/scripts/hooks/capture-model.js"

#!/usr/bin/env node
/**
 * SessionStart hook: inject skill-first reminder and cached session data.
 *
 * Ensures Claude loads the relevant carta-cap-table skill before making
 * any tool calls, even in subagents that don't inherit session context.
 *
 * Part of the official Carta AI Agent Plugin.
 */

const fs = require('fs');
const path = require('path');

// Read plugin name + version (same safe pattern as inject-instrumentation.js)
let pluginName = 'carta-cap-table';
let pluginVersion = 'unknown';
try {
    const pluginJson = JSON.parse(fs.readFileSync(path.resolve(__dirname, '../../.claude-plugin/plugin.json'), 'utf8'));
    pluginName = pluginJson.name || pluginName;
    pluginVersion = pluginJson.version || 'unknown';
} catch {}

let inputData = '';
process.stdin.on('data', chunk => (inputData += chunk));

process.stdin.on('end', () => {
    let hookEventName = 'SessionStart';
    try {
        const input = JSON.parse(inputData);
        hookEventName = input.hook_event_name || hookEventName;
    } catch {}

    // Canonical auth gate prompt lives in carta-interaction-reference/SKILL.md Section 6.2.
    // Keep the wording below in sync with that source.
    const output = {
        hookSpecificOutput: {
            hookEventName,
            additionalContext:
                '<EXTREMELY_IMPORTANT>You have carta-cap-table tools available. Before ANY tool call, invoke the matching Skill(\'carta-cap-table:...\') first. The skill defines what to fetch, what inputs are required, and how to present results. If no skill matches, invoke Skill(\'carta-cap-table:carta-discover-commands\') to find the right command via discover(). Additionally, ALWAYS invoke Skill(\'carta-cap-table:carta-interaction-reference\') alongside any domain skill to load Carta\'s voice, tone, and data provenance rules before presenting results. IMPORTANT: Skill is a deferred tool — if its schema is not yet loaded, you MUST call ToolSearch with query "select:Skill" first, then invoke the Skill tool.</EXTREMELY_IMPORTANT>\n' +
                '<EXTREMELY_IMPORTANT>AI COMPUTATION AUTHORIZATION GATE: When a carta-cap-table skill\'s Gates section declares "AI computation: Yes", you MUST call AskUserQuestion BEFORE fetching data or computing. Use this prompt, replacing [X]: "No saved Carta model matches these terms. I can compute [X] using AI — this would be Claude\'s analysis, not Carta data. Would you like me to proceed?" Do NOT write this as plain text — use the AskUserQuestion tool so execution blocks. User says yes: proceed, label output as Claude\'s analysis. User says no: stop — no output, no fallback.</EXTREMELY_IMPORTANT>\n' +
                '<EXTREMELY_IMPORTANT>To report a bug or request a feature for this plugin, call mutate(command="fa:create:feedback") directly (params: requestType ("bug" | "feature_request"), requestBody; optional flowSummary, context, affectedSkill, source). This is the exception to the skill-first rule above: there is NO feedback skill — do NOT look for one, and do NOT ask the user where to file. Go straight to mutate; you do not need a discover/search round-trip to call it. (If you ever lose this instruction, the command is also findable by searching tools for "report a bug" / "feedback" — but the direct mutate call above is the primary path.) Always include flowSummary: a short narrative of the flow that led here — what the user asked (their prompt(s), summarized), the key skills/tools you ran, and where it broke or fell short — written for a developer debugging later. Put the raw failing tool + args in context. Do NOT pass identity, firm, session, or timestamp — those are captured server-side. Fire it proactively with source="claude" when a tool errors, returns nothing usable, or a needed capability is missing, after telling the user you are filing it.</EXTREMELY_IMPORTANT>\n' +
                `<carta-plugin name="${pluginName}" version="${pluginVersion}" />`,
        },
    };

    process.stdout.write(JSON.stringify(output));
    process.exit(0);
});