add-problem

Add metadata for a newly-solved problem to the configured metadata file. Skip the manual research grind.

Step 0 — Load config

Look for dsa-coach.config.json in the current working directory. If present, parse it and use its values. For any missing key, fall back to the documented default below. If the file is absent entirely, use all defaults.

Config keys + defaults:

• language (default "ts") — solution file extension.
• solutionDirs (default { "leetcode": "leetcode", "geeksforgeeks": "geeksforgeeks", "hackerrank": "hackerrank", "interviewbit": "interviewbit" }) — platform key → folder name.
• metaFile (default "data/problems-meta.json") — metadata file to read/write.
• statsCommand (default "npm run stats") — optional stats regen command; skip gracefully if null/absent.

Throughout this skill, metaFile means the configured metadata path and the solutionDirs values are the folder names to match against.

Step 1 — Resolve target file

$ARGUMENTS is the repo-relative path to the newly-added solution file (e.g. leetcode/704_binary_search.ts).

If empty: run git status --short scoped to the directory holding metaFile plus every solutionDirs value (e.g. git status --short data/ leetcode/ geeksforgeeks/ hackerrank/ interviewbit/, using the configured folders) to find new untracked solution files with the configured language extension. If multiple, ask which one. If one, use it.

If the file already has an entry in metaFile under problems: warn, ask whether to refresh metadata or abort.

Step 2 — Detect platform + identifier

From the path, match the leading folder against the solutionDirs values (the examples below use the defaults; substitute the configured folder names and language extension):

• <leetcode-dir>/<NNN>_<slug>.<language> → LeetCode, problem ID = <NNN>
• <geeksforgeeks-dir>/<topic>/<slug>.<language> → GFG, slug = filename
• <hackerrank-dir>/<slug>.<language> → HackerRank, slug = filename
• <interviewbit-dir>/<slug>.<language> → InterviewBit

Read the file's first ~20 lines. If a // Problem link - <url> comment exists, it's authoritative — use that URL.

Step 3 — Fetch official metadata

LeetCode

• WebSearch: LeetCode <NNN> "<inferred title>" difficulty topics tags
• Extract: official title (exact casing), difficulty, topic tags
• LC blocks WebFetch (403). Use search snippets — they typically include difficulty + tags.
• Cross-check: WebFetch https://github.com/doocs/leetcode/blob/main/solution/<padded-range>/<NNN>...README_EN.md for tags.

GeeksforGeeks

• GFG is a JS-rendered SPA — WebFetch returns shell only. Skip direct fetch.
• WebSearch: GeeksforGeeks "<problem name>" practice difficulty topic tags
• Use the https://www.geeksforgeeks.org/problems/<slug>/1 URL when available.
• If problem comment in source file references gfg-160 batch, official title comes from that URL slug.

HackerRank

• WebFetch the /challenges/<slug>/problem page directly — usually returns difficulty + subdomain.

Step 4 — Pull LeetCode company frequencies (LC only)

Run via Bash:

tmp=$(mktemp -d)
ID=<problem-id>
COMPANIES="amazon google microsoft meta apple bloomberg uber adobe netflix tiktok oracle goldman-sachs nvidia salesforce linkedin atlassian airbnb"
for co in $COMPANIES; do
  curl -sL "https://raw.githubusercontent.com/snehasishroy/leetcode-companywise-interview-questions/master/$co/all.csv" \
    -o "$tmp/$co.csv"
done
for co in $COMPANIES; do
  freq=$(awk -F',' -v id="$ID" 'NR>1 && $1==id{print $NF}' "$tmp/$co.csv" 2>/dev/null)
  if [ -n "$freq" ]; then echo "$co $freq"; fi
done

Selection rule: companies with frequency ≥ 50%. If more than 5 qualify, take top 5 by frequency. If fewer than 5 qualify but problem has any, include all.

Capitalize properly when writing to JSON: meta → Meta, tiktok → TikTok, nvidia → NVIDIA, goldman-sachs → Goldman Sachs, linkedin → LinkedIn, otherwise Title Case.

For GFG / HackerRank / InterviewBit: leave companies: []. The aggregator above is LC-only and GFG company tags are gated behind their JS-rendered SPA — don't guess.

Step 5 — Determine category and patterns

• category = single primary topic. Heuristics: Array, String, Sorting, Linked List, Tree, Graph, Dynamic Programming. Pick the dominant one — usually matches the source folder for GFG.
• patterns = the official topic tags array from Step 3, verbatim. Don't invent tags. If source uses a famous algorithm name not in LC tags (e.g. "Boyer-Moore"), do NOT include it — stick to platform-official tags only.

If a new pattern appears that's not in the existing corePatterns array in metaFile, add it there too.

Step 6 — Write the entry

Read metaFile. If it does not exist, self-heal: create it with the skeleton { "corePatterns": [], "problems": {} } before inserting. Use Edit to insert the new entry into problems (keep the file path key as <platform>/<file> matching $ARGUMENTS exactly).

Entry shape:

"<file-path>": {
  "title": "<official title with exact casing>",
  "difficulty": "<Easy|Medium|Hard>",
  "category": "<primary topic>",
  "patterns": ["<official tag>", ...],
  "companies": ["<Company>", ...]
}

After write, validate JSON. Use the configured metaFile path:

python3 -c "import json; json.load(open('<metaFile>'))" && echo OK

Step 7 — Regenerate stats

If statsCommand is configured (non-null), run it:

<statsCommand>

Confirm the script prints the updated problem count.

If statsCommand is null or absent, print "stats command not configured — skipping" and continue.

Step 8 — Report

Print a 5-line summary:

• Title + source link
• Difficulty / category
• Patterns chosen (and source — LC official, GFG search, etc.)
• Companies (and freq cutoff applied)
• Anything skipped or uncertain — be honest, don't paper over gaps

$ARGUMENTS