AgentSkillsCN

LoggedAnalysis

分析指定日期的所有已记录网页搜索文件，并生成一份简洁的评估。

SKILL.md
--- frontmatter
name: LoggedAnalysis
description: Analyse all logged web search files for a given date and produce a concise assessment.
capabilities:
  - filesystem:read
  - filesystem:write:workspace
metadata: {"openclaw":{"requires":{"bins":["python"]}}}

This skill reads previously logged search files from the datastore and generates 3 high-level conclusions based on a user-provided assessment criteria.

It is used to summarize or interpret a day’s research activity.

When to use

Use this skill when the user asks to:

  • •analyse searches from a specific day
  • •summarize research activity
  • •identify trends in logged searches
  • •draw conclusions from prior browsing/search logs
  • •review what was learned on a given date

Do NOT use this skill for live web search. It only reads existing logs.

Inputs

  • •

    date (string, required) Format: YYYY/MM/DD or YYYY-MM-DD
    The day of logs to analyse.

  • •

    criteria (string, required) A short instruction describing how the logs should be assessed. Examples:

    • •"Check for trends"
    • •"Summarise main themes"
    • •"Identify recurring interests"
  • •

    name (string, optional) Label used for the output filename.

Execution

This skill runs a Python script.

Command:

python LoggedAnalysis.py --date "<date>" --criteria "<criteria>" --name "<name>"

Environment:

OPENCLAW_OUTPUT_ROOT optionally points to the datastore root. If not set, the skill will use the workspace root when a datastore folder exists, otherwise it falls back to C:\Users<user>.openclaw\data.

The script reads from:

<root>/datastore/YYYY/MM/DD/

and writes to:

<root>/analysis/YYYY/MM/DD/

Behavior

The skill:

  1. •Loads all .txt files from: datastore/YYYY/MM/DD/
  2. •Builds a compact corpus of the day’s logs
  3. •Applies the assessment criteria
  4. •Produces exactly 3 conclusions
  5. •Writes a JSON analysis file to: analysis/YYYY/MM/DD/

It returns:

  • •the 3 conclusions
  • •the output file path

Output

JSON object containing:

  • •date
  • •conclusions (array of 3 strings)
  • •output_path
  • •diagnostics (paths + file counts)

Examples

User:
"Analyse 2026/02/16 logs and check for trends"

→ Mechanical form:

LoggedAnalysis --date "2026/02/16" --criteria "Check for trends"

User:
"Summarise yesterday’s searches"

→ Convert to date, then call:

LoggedAnalysis --date "<resolved YYYY/MM/DD>" --criteria "Summarise main themes"

Notes

  • •This skill does not access the internet
  • •It only reads existing logs
  • •It is deterministic and file-backed
  • •It is safe to run repeatedly