scriptedalchemy/reverie-mcp/agentic-history-search

Agentic History Search

Use this skill for project-memory questions, such as:

• "what are the test commands?"
• "how did we implement xyz and why?"
• "what does function x do?"
• "when did we fix this bug before?"

Hard guardrails

1. Use conversation history evidence only.
2. If evidence is weak, explicitly say so.
3. Always return citations when answering.
4. Keep scope project-specific unless user asks otherwise.

Execution

Use the launcher (compiled bundle required):

bash skills/agentic-history-search/scripts/run-history-search.sh --mode query_past --query "<user question>"

Recent sessions helper mode:

bash skills/agentic-history-search/scripts/run-history-search.sh --mode recent_sessions --limit 10

The launcher requires skills/agentic-history-search/dist/history-search.mjs. This skill is expected to ship with prebuilt artifacts in dist/.

Output expectations

For query_past, include:

• concise answer grounded in history
• citations with provider + snippet + source ids when available
• confidence + insufficientEvidence flag
• provider diagnostics

For recent_sessions, include:

• ordered recent sessions
• provider diagnostics

References

• Detailed output contract: references/OUTPUT_SCHEMA.md
• Provider notes and capability behavior: references/PROVIDER_COMPATIBILITY.md
• Usage patterns and examples: references/USAGE_PATTERNS.md