cesnimda/jobtrackingapp
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

feat(ai): prompt-injection delimiters + synonym-aware match scoring
CI and Deploy / test (pull_request) Successful in 2m6s
Details
CI and Deploy / deploy (pull_request) Has been skipped
Details

Browse Source 
Wave 4 hardening. Wrap untrusted CV/job-description/instruction text
in tools/summarizer prompts with explicit delimiters and an
ignore-embedded-instructions rule, since JD text, recruiter emails,
and free-text candidate background all flow into rewrite/normalize
prompts unescaped today.

Match score previously normalized synonyms (JS/Kubernetes/K8s/etc)
only when scanning the job posting, not when checking the CV corpus,
so a CV using an abbreviation the job spelled out never matched.
SkillTagger.MatchesTag reuses the same synonym regex for both sides.
This commit is contained in:
cesnimda
2026-07-11 23:06:52 +02:00
parent fc62a659ef
commit 67ee3d7274
4 changed files with 64 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
JobTrackerApi/Services/JobCvMatchService.cs
+17 -5
View File
@@ -5,8 +5,9 @@ using JobTrackerApi.Services.JobImport;
namespace JobTrackerApi.Services
{
/// <summary>One keyword drawn from the job posting and whether the CV covers it.</summary>
public sealed record MatchKeyword(string Keyword, int Weight, bool InTitle, bool Matched);
/// <summary>One keyword drawn from the job posting and whether the CV covers it.
/// IsCuratedTag marks keywords sourced from SkillTagger, whose synonym regex is reused for CV matching.</summary>
public sealed record MatchKeyword(string Keyword, int Weight, bool InTitle, bool Matched, bool IsCuratedTag = false);
/// <summary>How many of the matched keywords appear in a given CV section.</summary>
public sealed record MatchSectionCoverage(string Section, int Matched, int Total);
@@ -80,8 +81,17 @@ namespace JobTrackerApi.Services
.ToDictionary(kvp => kvp.Key, kvp => Normalize(kvp.Value), StringComparer.OrdinalIgnoreCase);
var fullCorpus = string.Join(" \n ", sectionCorpora.Values);
// Raw (non-normalized) text for curated tags, whose synonym regex needs real word boundaries/punctuation.
var rawSections = cvSections
.Where(kvp => !string.IsNullOrWhiteSpace(kvp.Value))
.ToDictionary(kvp => kvp.Key, kvp => kvp.Value, StringComparer.OrdinalIgnoreCase);
var rawCorpus = string.Join(" \n ", rawSections.Values);
var evaluated = keywords
.Select(k => k with { Matched = CorpusContains(fullCorpus, k.Keyword) })
.Select(k => k with
{
Matched = k.IsCuratedTag ? SkillTagger.MatchesTag(k.Keyword, rawCorpus) : CorpusContains(fullCorpus, k.Keyword),
})
.ToList();
var totalWeight = evaluated.Sum(k => k.Weight);
@@ -103,7 +113,9 @@ namespace JobTrackerApi.Services
var sectionCoverage = sectionCorpora
.Select(section => new MatchSectionCoverage(
section.Key,
evaluated.Count(k => CorpusContains(section.Value, k.Keyword)),
evaluated.Count(k => k.IsCuratedTag
? SkillTagger.MatchesTag(k.Keyword, rawSections.GetValueOrDefault(section.Key))
: CorpusContains(section.Value, k.Keyword)),
evaluated.Count))
.Where(sc => sc.Total > 0)
.OrderByDescending(sc => sc.Matched)
@@ -129,7 +141,7 @@ namespace JobTrackerApi.Services
foreach (var tag in SkillTagger.Detect(combined))
{
var inTitle = TitleContains(jobTitle, tag);
byKey[tag] = new MatchKeyword(tag, CuratedTagWeight + (inTitle ? TitleBonus : 0), inTitle, false);
byKey[tag] = new MatchKeyword(tag, CuratedTagWeight + (inTitle ? TitleBonus : 0), inTitle, false, IsCuratedTag: true);
}
// 2) Salient posting terms: frequency-ranked content words from the description.