Command History Artifact: Review and Improvements

[dkrupnic]

Hi,
I would like to share some findings about the history artifact (files → shell → history) and suggest possible improvements.

Current behavior
Currently, the artifact script scans only the user home directory (~/) with max_depth: 4 and checks files matching the following patterns:

name_pattern: [".*_history", ".*history", ".cosh_history", ".dash_history", ".esh_history", ".lesshst", ".nash_history", ".sash_history", ".scsh_history", ".xonsh_history", ".zhistory", "*.historynew", "fish_history"]
    max_depth: 4

Shell coverage analysis
I did a short research of shells that can be considered relevant today:

Shell / Tool	Detection type	Comment
Any shell with custom history file path	Not supported reliably	No env/config parsing (HISTFILE, XDG_*). Only files under ~/ are scanned. Custom paths outside home are missed; inside home they may be detected only if filename matches wildcards.
Bash	Wildcard match	.bash_history is not explicitly listed, but matched by .*_history / .*history.
Ksh93 / mksh	Wildcard match	.sh_history, .ksh_history not explicitly listed; detected via wildcards.
Zsh	Wildcard match	.zsh_history not explicitly listed; detected via wildcards.
less (pager)	Explicit match	.lesshst is explicitly listed (tool history, not shell).
Fish	Explicit match	fish_history explicitly listed; default path ~/.local/share/fish/fish_history is within depth.
Xonsh	Explicit + wildcard (partial)	.xonsh_history explicitly listed (old versions). Newer JSON/SQLite in ~/.local/share/xonsh/ may be matched only if filename contains history.
Tcsh / Csh	Wildcard match	.history, .tcsh_history matched via wildcards.
Dash	Explicit + wildcard	.dash_history explicitly listed, but would also be matched by wildcards. Non-standard; appears only if user enables history manually.
Nushell (nu)	Not supported reliably	Uses its own data directory; may be missed unless filename accidentally matches wildcard under ~/.
Elvish	Wildcard match	Commonly ~/.elvish/history; filename history matches .*history. Structured/binary format.
PowerShell (pwsh) on Linux	Not supported reliably	History at ~/.local/share/powershell/PSReadLine/ConsoleHost_history.txt is deeper than max_depth: 4 and filename does not match patterns.
Oil (osh/ysh)	Not supported reliably	Uses custom mechanism; will be detected only if it writes under ~/ with history in filename.
Ion	Not supported reliably	Uses custom mechanism; will be detected only if it writes under ~/ with history in filename.

Questions about the current pattern:

1. The current pattern relies on wildcard values (e.g. .history, .*_history, *.historynew), which can cause false positives. Why both wildcards and explicit values were used that overlap with them?
2. Is cosh_history a typo? Should it be csh_history instead?
3. .tcsh_history is the default history file for tcsh. Should it be added, especially if wildcard patterns will be removed?
4. Why *.historynew was included? In which real cases does this occur?
   As far as I can see, only zsh and bash use rotation mechanisms, and they typically produce files like:
   .zsh_history.new, .zsh_history.tmp
   .bash_history~, .bash_history.[0-9]
   The glob *.historynew would not match any of these.

Suggestions

1. Rework the current name_pattern list to make it more explicit, safe, and predictable (avoid overly broad wildcards).
2. Consider dynamic detection of history files, for example:
   - using environment variables (HISTFILE, XDG_DATA_HOME, etc.),
   - or parsing shell configuration files.

This would improve shell coverage, support modern shells, and allow detection of custom history paths.

Thanks for your time and for maintaining this project. I hope these findings and suggestions are helpful.