Menu
Inspect this agent's stored JSONL session transcripts — list, read, search, or recall recent turns.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
| name | Session Logs |
| description | Inspect this agent's stored JSONL session transcripts — list, read, search, or recall recent turns. |
| requires | {"bins":[],"env":[]} |
Use this skill when the agent needs to look at its own prior conversations:
"what did I tell you yesterday?", "search whether we ever discussed X",
"show me the last 5 things from this session". Reads the JSONL transcripts under
transcripts_dir (Phase 10.4 writes them).
memory tool
(vector/keyword) instead; this one is per-session raw transcriptSingle tool session_logs with action dispatch:
list_sessionslimit (integer, optional, 1–500, default 50) — most-recent firstReturns {sessions: [{session_id, agent_id, source_plugin, entry_count, first_timestamp, last_timestamp}, ...]}.
read_sessionsession_id (UUID, required)limit (integer, optional) — max entries returnedmax_chars (integer, optional, 20–4000, default 200) — truncate each content previewReturns {header, total_entries, returned, truncated, entries: [...]}.
searchquery (string, required) — case-insensitive substring matchlimit (integer, optional, default 50)max_chars (integer, optional) — preview length per hitReturns {hits: [{session_id, timestamp, role, source_plugin, preview}, ...]}.
recentsession_id (UUID, optional) — defaults to the current session when omittedlimit (integer, optional, default 10)max_chars (integer, optional)Returns {session_id, count, entries: [...]}.
recent for "what did I just say" questions — cheap and
enough for short windows.search for "did we ever discuss X" queries; pair with
list_sessions if you need to know which session to then read_session.max_chars — full transcripts can blow the LLM
context window fast.transcripts_dir is not configured, the tool returns
{ok: false, error: "transcripts_dir is not configured..."} — ask the
operator to set agents[].transcripts_dir in agents.yaml.memory tool with mode: "vector".Bounded auto-iteration for a prompt with explicit stop predicates.
Memory hygiene workflow to classify, deduplicate, and promote durable knowledge across local memory layers.
Bounded simplification pass for a file or hunk that preserves behavior while reducing complexity.
Turn a repeatable workflow into a reusable local SKILL.md with explicit steps, criteria, and safe defaults.
Bounded auto-debug loop for failing build/test commands with evidence-first diagnosis.
Safely update Nexo config files with read-before-write merges, schema-aware mapping, and reload/restart awareness.