Overview

This is an item-by-item list of which AI agent usage logs (OpenTelemetry) Skill Assessment receives, what it uses them for, and how long it keeps them. For an overview of masking, temporary storage, and post-evaluation retention, see How Skill Assessment Handles Measurement Logs.

  • AI agents covered: Claude Code and Codex CLI. If you use GitHub Copilot, the items it outputs are different, so please contact your customer success representative.
  • What each column shows:
Column Meaning
Sent Whether it is sent from your machines to HireRoo. "○" means it is sent when the tool outputs it
Used Whether HireRoo uses it for evaluation or aggregation. Items marked "×" are discarded on receipt and never stored
Temporarily stored Whether it is stored temporarily, in masked form, until evaluation finishes (see How Skill Assessment Handles Measurement Logs)
Retained after evaluation Whether it is kept after evaluation finishes (see How Skill Assessment Handles Measurement Logs)

1. Settings Changed on the Machine

The measurement CLI (hireroo spider setup) changes only the following settings.

  • How to revert: Running hireroo spider uninstall restores the changed settings to their values before installation.
  • Destination: https://spider-collector.hireroo.io/otel (HTTPS)
  • Authentication: A token issued per user is sent in an HTTP header.
Tool Setting changed Enabled Not enabled
Claude Code env in ~/.claude/settings.json Logs (events) / prompt text (OTEL_LOG_USER_PROMPTS) / tool details (OTEL_LOG_TOOL_DETAILS) / response text (OTEL_LOG_ASSISTANT_RESPONSES)* Metrics, Traces, API Body
Codex CLI [otel] in ~/.codex/config.toml Logs / prompt text (log_user_prompt = true) Response text (log_agent_responses), Traces, Metrics (HireRoo is not set as the destination)

* Sending of response text is enabled in preparation for future features. It is not currently used for evaluation; it is discarded on receipt and not stored temporarily (see 2-4).

HireRoo's endpoint does not accept Metrics. Traces from Claude Code and Codex are discarded without being read, even if received.

2. Claude Code Items

2-1. Common Session and User Attributes

Item Sent Used Temporarily stored Retained after evaluation
session.id ○ ○ ○ ○
ccr.session.id ○ × × ×
app.version ○ × (the resource's service.version is used instead) × ×
app.entrypoint ○ × × ×
user.id ○ × × ×
user.email ○ × × ×
user.account_uuid ○ × × ×
user.account_id ○ × × ×
organization.id ○ × × ×
terminal.type ○ × × ×
vcs.repository.url.full ○ × × ×
vcs.owner.name ○ × × ×
vcs.repository.name ○ × × ×
vcs.provider.name ○ × × ×
Additional attributes via OTEL_RESOURCE_ATTRIBUTES Not set by HireRoo × × ×

Attributes that identify users or organizations (user.*, organization.id, host.name) are excluded from reading on receipt.

2-2. Metrics (claude_code.*)

None of session.count, lines_of_code.count, pull_request.count, commit.count, cost.usage, token.usage, code_edit_tool.decision, or active_time.total are sent (because sending Metrics is not enabled).

2-3. Generative AI Call Events

Event Item Sent Used Temporarily stored Retained after evaluation
claude_code.api_request model ○ ○ ○ ○
input_tokens / output_tokens / cache_read_tokens / cache_creation_tokens ○ ○ ○ ○
cost_usd ○ ○ ○ ○
duration_ms ○ ○ (used to calculate start and end times) ○ ○ (as start and end times)
query_source ○ ○ (used to determine the call type) ○ ○ (as the call type)
cost_usd_micros, request_id, client_request_id, speed, effort, caller attributes other than query_source (agent.name, skill.name, plugin.name, marketplace.name, mcp_server.name, mcp_tool.name) ○ × × ×
claude_code.api_error Occurrence of the event ○ ○ (counted as a failure) ○ ○ (as success or failure)
error, status_code, attempt and other attributes ○ × × ×
claude_code.api_refusal All ○ × × ×
claude_code.api_retries_exhausted All ○ × × ×

2-4. User Interaction Events

Event Item Sent Used Temporarily stored Retained after evaluation
claude_code.user_prompt prompt (prompt text) ○ ○ ○ (masked) Character count, and excerpts used as the basis for the evaluation (masked; see How Skill Assessment Handles Measurement Logs)
prompt_length ○ × (HireRoo recounts the length after masking) × ×
message.uuid, command_name, command_source ○ × × ×
claude_code.assistant_response response (response text), response_length, and others ○ × (not currently used) × ×

2-5. Tool and MCP Events

Event Item Sent Used Temporarily stored Retained after evaluation
claude_code.tool_result tool_name, tool_use_id, success, duration_ms ○ ○ ○ ×
tool_input / tool_parameters (tool arguments, including file paths, commands, and edits) ○ ○ ○ (masked) × (used only for counts by language and aggregation of MCP server and tool names)
mcp_server_scope ○ ○ ○ ○ (as aggregated MCP usage)
error_type, error, decision_type, decision_source, tool_input_size_bytes, tool_result_size_bytes, vcs.ref.head.* ○ × × ×
claude_code.tool_decision tool_name, tool_use_id, decision, source ○ ○ ○ ×
tool_source ○ × × ×
claude_code.mcp_server_connection All ○ × × ×

2-6. Plugin, Skill, and Permission Events

Event Item Sent Used Temporarily stored Retained after evaluation
claude_code.skill_activated skill.name, skill.source ○ ○ ○ ○ (as aggregated skill usage)
invocation_trigger ○ ○ ○ ×
skill.kind ○ × × ×
plugin_loaded, plugin_installed, hook_registered, permission_mode_changed, auth, at_mention, internal_error All ○ × × ×

2-7. API Body and Traces

Item Sent
claude_code.api_request_body / api_response_body Not sent (OTEL_LOG_RAW_API_BODIES is not enabled)
Traces (interaction / llm_request / tool / hook spans) Not sent (sending Traces is not enabled)

2-8. Common Time Attributes

Item Sent Used Temporarily stored Retained after evaluation
event.timestamp ○ ○ ○ ○ (as start and end times)
event.sequence ○ × × ×

2-9. Other Items Used by HireRoo

Item Purpose Temporarily stored Retained after evaluation
event.name Determining the event type × ×
prompt.id Grouping by prompt ○ ×
The resource's service.name / service.version Determining the tool and version ○ ○

3. Codex CLI Items

3-1. Common Session Attributes

Item Sent Used Temporarily stored Retained after evaluation
conversation_id ○ ○ ○ ○
model ○ ○ ○ ○
account_id, user_id, account_email ○ × × ×
turn.id, agent_name, slug, service_tier, model_reasoning_effort, auth_mode, originator, product_sku, session_source, terminal_type ○ × × ×
service_name, app.version ○ × (the resource's service.name / service.version are used instead) × ×

3-2. Metrics (codex.*) and Common Metric Tags

HireRoo is not set as a Metrics destination, so they are not sent to HireRoo.

3-3. Events

Event Item Sent Used Temporarily stored Retained after evaluation
codex.tool_result tool_name, call_id, success, duration_ms ○ ○ ○ ×
arguments (tool arguments) ○ ○ ○ (masked) × (used only for counts by language and aggregation of skill names)
mcp_server ○ ○ ○ ○ (as aggregated MCP usage)
output (preview of tool output) ○ × × ×
tool_namespace, output_truncated, tool_result_seq, product_sku, agent_name, mcp_server_origin ○ × × ×
Attributes attached only to traces (arguments_length and others) Not sent — — —
codex.api_request Success or failure ○ ○ ○ ○
codex.agent_response All Not sent (log_agent_responses is not enabled) — — —
codex.skill_invocation All ○ × × ×
codex.guardian_assessment All Not sent (log_guardian_assessments is not enabled) — — —
codex.retry, codex.turn_cost, codex.websocket_request, codex.tool_call family, codex.agent_communication, codex.network_proxy.*, codex.http_client.*, codex.exec_server.*, and others All ○ × × ×

3-4. Other Items Used by HireRoo

Event Item Purpose Temporarily stored Retained after evaluation
codex.user_prompt prompt (prompt text), model Evaluation ○ (masked) Character count, and excerpts used as the basis for the evaluation (masked; see How Skill Assessment Handles Measurement Logs)
codex.sse_event (response.completed) Token counts for input, cache reads, cache writes, and output Aggregation of usage ○ ○
codex.tool_decision call_id, tool_name, decision, source Evaluation ○ ×
Common event.timestamp, the resource's service.name / service.version Timestamps, and determining the tool and version ○ ○

HireRoo calculates cost based on publicly available price lists.