Skill3.8k estrellas del repoactualizado yesterday
session-investigator
Session Investigator examines `.fast-agent/sessions/` files to diagnose session failures and tool execution issues. Use it when sessions terminate unexpectedly, when debugging stuck tool loops, when tracing correlations between sub-agent and main agent histories, or when analyzing message timing and flow. This skill provides commands to inspect session metadata, correlate tool calls with results by ID, review message sequences, and identify common failure patterns like missing tool results or broken handoffs.
Instalar en Claude Code
Copiargit clone --depth 1 https://github.com/evalstate/fast-agent /tmp/session-investigator && cp -r /tmp/session-investigator/examples/hf-toad-cards/skills/session-investigator ~/.claude/skills/session-investigatorDespués abre una sesión nueva de Claude Code; el skill carga automáticamente.
Definición
SKILL.md
# Session Investigator
Diagnose fast-agent session issues by examining session and history files.
## Session Directory Structure
Sessions are stored in `.fast-agent/sessions/<session-id>/`:
```
2601181023-Kob2h3/
├── session.json # Session metadata
├── history_<agent>.json # Current agent history
└── history_<agent>_previous.json # Previous save (rotation backup)
```
Session IDs encode creation time: `YYMMDDHHMM-<random>` (e.g., `2601181023` = 2026-01-18 10:23).
## Key Files
### session.json
```json
{
"name": "2601181023-Kob2h3",
"created_at": "2026-01-18T10:23:24.116526",
"last_activity": "2026-01-18T10:39:42.873467",
"history_files": ["history_dev_previous.json", "history_dev.json"],
"metadata": {
"agent_name": "dev",
"first_user_preview": "is it possible to override..."
}
}
```
### history\_<agent>.json
```json
{
"messages": [
{
"role": "user|assistant",
"content": [{"type": "text", "text": "..."}],
"tool_calls": {"<id>": {"method": "tools/call", "params": {"name": "...", "arguments": {}}}},
"tool_results": {"<id>": {"content": [...], "isError": false}},
"channels": {
"fast-agent-timing": [{"type": "text", "text": "{\"start_time\": ..., \"end_time\": ..., \"duration_ms\": ...}"}],
"fast-agent-tool-timing": [{"type": "text", "text": "{\"<tool_id>\": {\"timing_ms\": ..., \"transport_channel\": ...}}"}],
"reasoning": [{"type": "text", "text": "..."}]
},
"stop_reason": "endTurn|toolUse|error",
"is_template": false
}
]
}
```
## Investigation Commands
### Basic inspection
```bash
# Message count
jq '.messages | length' history_dev.json
# Last N messages overview
jq '.messages[-5:] | .[] | {role, stop_reason, has_tool_calls: (.tool_calls != null), has_tool_results: (.tool_results != null)}' history_dev.json
# View specific message
jq '.messages[227]' history_dev.json
```
### Tool call correlation
Tool calls and results are linked by correlation ID. Valid pattern: assistant with `tool_calls` → user with matching `tool_results`.
```bash
# Check tool call/result pairing
jq '.messages[-10:] | to_entries | .[] | {
index: .key,
role: .value.role,
tool_calls: (if .value.tool_calls then (.value.tool_calls | keys) else [] end),
tool_results: (if .value.tool_results then (.value.tool_results | keys) else [] end)
}' history_dev.json
```
### Find specific tool calls
```bash
# Find all calls to a specific tool
jq '.messages | to_entries | .[] |
select(.value.tool_calls != null) |
select(.value.tool_calls | to_entries | .[0].value.params.name == "agent__ripgrep_search") |
{index: .key, timing: (.value.channels."fast-agent-timing"[0].text)}' history_dev.json
```
## Session Statistics
### LLM Call Stats
```bash
# Total LLM time and call count
jq '[.messages[] | select(.role == "assistant") |
select(.channels."fast-agent-timing") |
.channels."fast-agent-timing"[0].text | fromjson | .duration_ms] |
{count: length, total_ms: add, avg_ms: (add/length), max_ms: max, min_ms: min}' history_dev.json
# LLM calls sorted by duration (slowest first)
jq '[.messages | to_entries | .[] |
select(.value.role == "assistant") |
select(.value.channels."fast-agent-timing") |
{index: .key, duration_ms: (.value.channels."fast-agent-timing"[0].text | fromjson | .duration_ms)}] |
sort_by(-.duration_ms) | .[0:10]' history_dev.json
```
### Tool Execution Stats
```bash
# All tool timings aggregated
jq '[.messages[] | select(.channels."fast-agent-tool-timing") |
.channels."fast-agent-tool-timing"[0].text | fromjson | to_entries | .[].value.timing_ms] |
{count: length, total_ms: add, avg_ms: (add/length), max_ms: max, min_ms: min}' history_dev.json
# Tool calls by name with timing
jq '[.messages | to_entries | .[] |
select(.value.tool_calls) |
(.value.tool_calls | to_entries | .[0]) as $tc |
{index: .key, tool: $tc.value.params.name,
llm_ms: (.value.channels."fast-agent-timing"[0].text | fromjson | .duration_ms)}] |
group_by(.tool) |
map({tool: .[0].tool, count: length, total_llm_ms: (map(.llm_ms) | add)}) |
sort_by(-.count)' history_dev.json
```
### Session Timeline
```bash
# Session duration from first to last timing
jq '.messages | [
(map(select(.channels."fast-agent-timing")) | first | .channels."fast-agent-timing"[0].text | fromjson | .start_time),
(map(select(.channels."fast-agent-timing")) | last | .channels."fast-agent-timing"[0].text | fromjson | .end_time)
] | {start: .[0], end: .[1], duration_sec: ((.[1] - .[0]) | round)}' history_dev.json
# Message rate over time (messages per minute estimate)
jq '{
messages: (.messages | length),
llm_calls: [.messages[] | select(.role == "assistant" and .channels."fast-agent-timing")] | length,
total_llm_ms: [.messages[] | select(.channels."fast-agent-timing") | .channels."fast-agent-timing"[0].text | fromjson | .duration_ms] | add,
total_tool_ms: [.messages[] | select(.channels."fast-agent-tool-timing") | .channels."fast-agent-tool-timing"[0].text | fromjson | to_entries | .[].value.timing_ms] | add
} | . + {llm_sec: (.total_llm_ms/1000), tool_sec: ((.total_tool_ms//0)/1000)}' history_dev.json
```
### Sub-agent Stats
```bash
# Sub-agent calls (tools starting with "agent__")
jq '[.messages | to_entries | .[] |
select(.value.tool_calls) |
(.value.tool_calls | to_entries | .[0]) as $tc |
select($tc.value.params.name | startswith("agent__")) |
{index: .key, agent: $tc.value.params.name,
llm_ms: (.value.channels."fast-agent-timing"[0].text | fromjson | .duration_ms)}] |
group_by(.agent) |
map({agent: .[0].agent, calls: length, total_ms: (map(.llm_ms) | add), avg_ms: ((map(.llm_ms) | add) / length)})' history_dev.json
```
## Common Failure Patterns
### Unanswered Tool Call
**Symptom**: API error "No tool output found for function call"
**Pattern**: History ends with `assistant` message having `tool_calls` and `stop_reason: "toolUse"`, followed by `user` message WITH